tsunami evacuation modelling - gitews · tsunami evacuation modelling ... planning using geo-...
TRANSCRIPT
INSTITUT FÜR GEOGRAPHIE AN DER UNIVERSITÄT REGENSBURG
Tsunami Evacuation Modelling
Development and application of a spatial information system
supporting tsunami evacuation planning in South-West Bali
VON: MATTHIAS MÜCK
REGENSBURG 2008
DIPLOMARBEIT ZUR ERLANGUNG DES DIPLOMGRADES (DIPL.- GEOGR. UNIV.)
AM INSTITUT FÜR GEOGRAPHIE DER UNIVERSITÄT REGENSBURG
JUNI 2008
DIE ARBEIT WURDE EINGEREICHT AM: INSTITUT FÜR GEOGRAPHIE, REGENSBURG
BETREUER: APL. PROF. DR. KURT KLEIN
DIE ARBEIT WURDE ANGEFERTIGT AM: DLR OBERPFAFFENHOFEN
DEUTSCHES FERNERKUNDUNGSDATENZENTRUM (DFD)
82230 WEßLING
I
Acknowledgements
First and foremost I offer my sincerest gratitude to my supervisor, Dr. Joachim Post from DLR,
who has supported me throughout my thesis with his knowledge and great interest. I attribute the
level of my thesis to his encouragement and effort. Without him this thesis would not have been
written in this way.
I gratefully thank Harald Spahn from GTZ- IS for his advice, supervision and guidance from the
very early stage of this research. I greatly benefited by his support during the field work in Bali.
His crucial contribution will be remembered as a positive example for an outstanding teamwork.
I am very grateful to Dr. Torsten Riedlinger for the extraordinary support and great patience
during the final stage of this research. His contribution to the layout design and formatting process
leaded to a crucial improvement of this thesis.
Many thanks go in particular to Prof. Dr. Kurt Klein for giving me the backing for this project. His
great encouragement during the whole process time (and also during my whole study period) was
essential for the success of this work.
Collective acknowledgements are also owed to my colleagues at DFD for the numerous advices
and great working atmosphere during the preparation time of the thesis.
It is also a great pleasure to pay tribute to my Indonesian friends and partners for the numerous
and crucial contributions. Life was great in Indonesia. My special thanks go to Erma Novitasari for
her unremitting support during the field work in Kuta. Many thanks to Andi from the shoe shop for
the lunch meetings and fruitful discussions about life.
With my sincere gratitude I pay tribute to my girlfriend Astrid for being unselfishly let her
intelligence, passion and ambitions collide with mine.
Where would I be without my family? My parents deserve special mention for their irreplaceable
support during my studies, for giving the fundament of my learning character and showing me the
joy of intellectual pursuit ever since I was a child. Words fail me to express my appreciation to my
mother who sincerely raised me with her caring and gently love, and stands behind me in difficult
times.
II
III
Summary
This research aims at developing a spatial information system supporting tsunami evacuation
planning using geo- information technology. A case study was conducted in a tsunami- prone area
in the South- West of Bali.
Because of the great earthquake in the northwest of Sumatra island, Indonesia, the devastating
tsunami of December 26, 2004 has become the most important research topic in tsunami research
and disaster mitigation efforts today. The main effort in minimising casualties in tsunami disaster is
to evacuate people from the hazard area before tsunami strikes by means of either horizontal or
vertical evacuation.
Within the thesis an assessment of people’s ability to evacuate is presented. This assessment
requires knowledge of probable tsunami sources and impact characteristics along the coast. To
address the spatial and temporal component of the population at risk, a population disaggregation
concept was developed to identify people’s movements between different functional urban sectors
for day and night- times. This includes people commuting into the study area and tourists.
Accessibility analysis on the basis of a Cost Weighted Distance (CWD) approach is utilised to
identify optimized evacuation routes from a defined initial location to the nearest evacuation
shelter. Robust multi-storey buildings which meet the general requirements of an evacuation
shelter building (ESB) as well as safe areas based on a tsunami inundation scenario were identified
in the study area. In addition, a criteria set for ESB’s has been worked out.
A final evacuation modelling is combining the accessibility modelling with the exposed population
in order to give a statement about the evacuation ability of the population at risk for different
evacuation times. Results of the modelling include evacuation bottlenecks based on shelter
capacities, population hotspots and available evacuation routes.
Based on one inundation scenario, the application of the evacuation modelling in the study area
shows that a wide majority of the exposed population is able to access an evacuation shelter
within an evacuation time of 25 min. However, a multi- scenario approach is required in order to
improve the identification of areas affected by tsunami impact. Central research expertises are the
requirement of detailed population data to ensure effective evacuation planning and the need to
allocate additional evacuation shelters particularly at locations with high population density. A
crucial point is the limited capacity of some evacuation shelter buildings in the study area, which
leads to an insufficient evacuation ability in some shelter catchment areas. Based on these results,
additional ESB planning can be recommended to decision makers.
By addressing the above- mentioned issues that greatly influence the evacuation process, the
research provides essential spatial information for local decision makers in tsunami- prone areas
based on a modelling approach transferable to other coastal areas. .
V
Ringkasan
Penelitian ini bertujuan untuk mengembangkan sebuah sistem informasi spasial yang mendukung
perencanaan evakusi dengan menggunakan tehnologi geoinformasi. Daerah rawan bencana
tsunami di Barat Daya Bali merupakan wilayah yang dipilih untuk studi kasus.
Sejak kejadian gempa besar di Barat Laut kepulauan Sumatera dan mengakibatkan tsunami 26
Desember 2004, topik penelitian di bidang tsunami dan pengurangan dampak bencana menjadi
suatu penelitian yang sangat penting. Upaya utama untuk meminimalkan korban dalam bencana
tsunami adalah evakusi orang di wilayah bencana sebelum tsunami terjadi baik secara horisontal
maupun vertikal. Penilaian kemampuan masyarakat untuk evakuasi telah disajikan dalam thesis
ini. Dalam penilaian ini membutuhkan informasi tentang peluang kejadian tsunami dan
karakteristik dampaknya di sepanjang pantai. Untuk menunjukkan komponen populasi orang yang
memiliki resiko secara spasial dan waktu, konsep penyebaran populasi orang dibangun dengan
mengidentifikasi pergerakan orang di dalam wilayah fungsional yang berbeda di sektor perkotaan
siang dan malam. Dalam konsep ini, orang yang masuk dan keluar di area studi seperti turis dan
kaum pendatang diperhitungkan.
Pendekatan analisis aksesibilitas dengan basis Cost Weighted Distance (CWD) digunakan untuk
mengidentifikasi rute evakuasi yang optimal dari suatu lokasi yang telah ditentukan ke tempat
evakuasi terdekat. Gedung bertingkat yang kuat sebagai salah satu syarat umum untuk gedung
sebagai tempat evakuasi dan daerah yang aman berbasiskan skenario tsunami inundasi
diidentifikasi dalam studi area dan penentuan kriteria untuk gedung sebagai tempat evakuasi telah
dikerjakan dengan baik.
Hasil akhir model evakuasi adalah mengkombinasikan model aksesibiltas dengan orang yang
terekspos untuk memberikan suatu pernyataan tentang kemampuan evakuasi dari suatu populasi
orang yang beresiko untuk waktu evakuasi yang berbeda. Hasil dalam model ini termasuk juga
evakuasi „bottleneck“ yang berdasarkan kapasitas tempat evakuasi, tempat orang berkumpul, dan
ketersediaan jalur evakuasi.
Berdasarkan pada satu skenario inundasi, model aplikasi di area studi menunjukkan secara umum
bahwa mayoritas orang yang terekspos dapat mencapai tempat evakuasi dengan waktu evakuasi
sekitar 25 menit. Bagaimanapun juga, pendekatan multi skenario dibutuhkan untuk menentukan
secara tepat daerah yang terkena dampak tsunami. Penelitian ini membutuhkan informasi orang
secara detail untuk menjamin perencanaan evakuasi yang efektif dan memerlukan tambahan
tempat evakuasi secara khusus untuk daerah populasi tinggi. Masalah penting adalah keterbatasan
kemampuan gedung sebagai tempat evakuasi di area studi, yang menyebabkan kemampuan
evakuasi yang tidak cukup di daerah-daerah tangkapan sebagai tempat evakuasi. Berdasarkan
hasil ini, penambahan perencanaan gedung sebagai tempat evakuasi dapat direkomendasikan
untuk para pengambil kebijakan. Dengan menunjukkan hal yang tersebut di atas, isu yang telah di
ungkap akan sangat mempengaruhi proses evakuasi. Penelitian ini telah menyediakan informasi
spasial yang penting untuk para pengambil keputusan lokal di daerah rawan tsunami berdasarkan
model yang dapat di transfer ke daerah pantai yang lain.
VII
Table of contents
Acknowledgements....................................................................................................................I
Summary ............................................................................................................................... III
Ringkasan ................................................................................................................................V
Table of contents .................................................................................................................. VII
List of abbreviations ................................................................................................................IX
List of figures ..........................................................................................................................XI
List of tables .......................................................................................................................... XV
1 Introduction........................................................................................................................ 1 1.1 Objectives and structure of the thesis ......................................................................... 2 1.2 State of research ....................................................................................................... 4 1.3 Approach................................................................................................................... 7
1.3.1 GIS for Natural Disaster Management .................................................................. 7 1.3.2 Tsunami evacuation planning and spatial decision support .................................... 7
2 Study area – Case study Kuta .............................................................................................. 9 2.1 Geographical position and political structuring ............................................................. 9 2.2 Economic factor tourism........................................................................................... 10 2.3 Tectonic structures .................................................................................................. 11 2.4 Tsunami generation and zonal activities in the region of Bali ...................................... 12
3 Methods ........................................................................................................................... 15 3.1 Data collection......................................................................................................... 15
3.1.1 Data requirement and data acquisition methods ................................................. 15 3.1.2 Data acquisition and field work preparation ........................................................ 16 3.1.3 Field work ........................................................................................................ 18
3.2 Hazard impact & exposed area ................................................................................. 23 3.2.1 Tsunami hazard assessment.............................................................................. 23 3.2.2 Degree of exposure and potential safe areas ...................................................... 27
3.3 Population modelling................................................................................................ 28 3.4 Accessibility modelling.............................................................................................. 33
3.4.1 Concept of Cost Weighted Distance (CWD)......................................................... 35 3.4.2 Evacuation shelter accessibility .......................................................................... 35
3.5 Evacuation modelling ............................................................................................... 39 3.5.1 Shelter area accessibility for day and night population scenario ........................... 39 3.5.2 Evacuation ability for day and night population scenarios .................................... 41
VIII
4 Result ...............................................................................................................................45 4.1 Data collection .........................................................................................................45 4.2 Hazard impact and exposed area...............................................................................51
4.2.1 Potential safe areas ...........................................................................................54 4.3 Population modelling ................................................................................................55 4.4 Accessibility modelling ..............................................................................................59 4.5 Evacuation modelling................................................................................................65
4.5.1 Shelter accessibility for day and night population scenario ...................................65 4.5.2 Evacuation ability for day and night population scenario ......................................68
5 Discussion.........................................................................................................................71 5.1 Data collection .........................................................................................................71 5.2 Hazard impact & exposed area ..................................................................................75 5.3 Population modelling ................................................................................................78 5.4 Accessibility modelling ..............................................................................................81 5.5 Evacuation modelling................................................................................................84 5.6 General discussion....................................................................................................90
6 Conclusion & Recommendations .........................................................................................93
7 Outlook.............................................................................................................................99
Bibliography..........................................................................................................................101
Expert interviews ..................................................................................................................104
Appendices ...........................................................................................................................105
IX
List of abbreviations
ADPC Asian Disaster Preparedness Centre
AWI Alfred Wegener Institute
BAKOSURTANAL Badan Koordinasi Survei dan Pemetaan Nasional (National Coordinating Agency for Surveys and Mapping)
BAPPEDA Badan Perencanaan Pembangunan (Regional Development Planning Board)
BMG Badan Meteorologi dan Geofisika (Meteorological and Geophysical Agency)
BPS Badan Pusat Statistik (Statistics Indonesia)
CWD Cost Weighted Distance
DLR German Aerospace Center
DSM Digital Surface Model
DTM Digital Terrain Model
ESB Evacuation Shelter Building
FEMA Federal Emergency Management Agency
GITEWS German-Indonesian Tsunami Early Warning System
GTZ- IS German Technical Cooperation – International Services
IHRA International Hotel and Restaurant Association
IOC Intergovernmental Oceanographic Commission
KESBANGLINMAS Kesatuan Bangsa dan Perlindungan Masyarakat (Office of National Unity and Community Protection)
LAPAN Lembaga Penerbangan dan Antariksa National (National Institute of Aeronautics and Space)
LIPI Lembaga Ilmu Pengetahuan Indonesia (Indonesian Institute of Sciences)
NGDC National Geophysical Data Centre
NGO Non- Governmental Organisation
NOAA National Oceanic and Atmospheric Administration
PMI Palang Merah Indonesia
(Indonesian Red Cross)
SDSS Spatial Decision Support System
SOP Standard Operating Procedure
TNI Armed Forces of Indonesia
UN- ISDR International Strategy for Disaster Reduction
XI
List of figures
Figure 1 Combination of GIS and dynamic simulation models (Source: own
illustration) ...................................................................................................................... 6
Figure 2 Geographical position of the study area (Source: own illustration -
Data basis: BPS CENSUS 2000; Map basis: Quickbird satellite image).................................. 9
Figure 3 Bali – Main source of income of most of the population (Source: own
illustration - Data basis: BPS PODES 2005; Map basis: BPS census 2000) .......................... 10
Figure 4 Seismotectonics of 6 Indonesian zones (A- F) (Source: LATIEF ET AL.
(2000), p.28)................................................................................................................. 12
Figure 5 Tsunami generation and propagation (Source:
http://www.important.ca/tsunami_causes.html) .............................................................. 12
Figure 6 Histogram of destructive earthquakes (1800 – 1999) and tsunamis
caused by earthquakes (1600 – 1999) by zone and by decade (Source: LATIEF ET
AL. (2000), p. 33, modified) ............................................................................................ 13
Figure 7 Identification of functional urban sectors in the study area, gained by
field work (Source: own illustration)................................................................................ 19
Figure 8 Decision tree to approximate the degree of hazard impact on land
(Source: Post et al. (2007), modified) ............................................................................. 24
Figure 9 (left) / 10 (right) Inundation scenario used for the evacuation
modelling in the study area (Source: own illustration - Data basis: Alfred
Wegener Institute (AWI), modified; Map basis: Indonesia overview, DLR)......................... 26
Figure 11 Requirements for safe areas (horizontal evacuation) (Source: own
illustration - Data basis: expert interview with Dr. Joachim Post, 15.04.2008).................... 27
Figure 12 Population disaggregation with a combination of land use and census
data (Source: own illustration - Data basis: BPS CENSUS 2000 and LAPAN; Map
basis: BPS CENSUS 2000)............................................................................................... 30
Figure 13 Population distribution concept for the study area considering
functional urban sectors (Source: own illustration - Data basis: own data
collection) .................................................................................................................... 32
Figure 14 Effect of different reference parameters for the population modelling
results in the study area (Source: own illustration)........................................................... 33
Figure 15 Computation of CWD between two adjacent cells (Source: ADPC
2007a, p.6) ................................................................................................................... 35
Figure 16 Process steps for the accessibility modelling approach (Source: own
illustration – Data basis: ADPC (2007a) ........................................................................... 38
XII
Figure 17 Relevant information to calculate people’s ability to evacuate (Source:
own illustration) .............................................................................................................39
Figure 18 Example for the shelter accessibility within 15 minutes (Source: own
illustration - Data basis: own calculation; Map basis: Quickbird satellite image) ..................40
Figure 19 Example for the evacuation ability within 15 minutes (Source: own
illustration - Data basis: own calculation; Map basis: Quickbird satellite image) ..................43
Figure 20 (left) /21 (right) Traffic volume (left) and road width (right) in the
study area (Source: own illustration – Data basis: own data collection; Map basis:
Quickbird satellite image)................................................................................................45
Figure 22 Functional urban sectors in the study area (Source: own illustration -
Data basis: own data collection; Map basis: Quickbird satellite image)...............................46
Figure 23 Critical facilities in the study area (Source: own illustration - Data
basis: own data collection; Map basis: Quickbird satellite image).......................................50
Figure 24 Tsunami hazard zonation of South Bali, based on historical records of
tsunami impact observation. From left: (a) probability of affectness regarding the
horizontal distance, (b) probability of affectness regarding elevation, (c)
probability of exposure as a combination of distance, elevation and surface
roughness (Source: own illustration - Data basis: NGDC; Map basis: Quickbird
satellite image)...............................................................................................................52
Figure 25 (left) / 26 (right) Hazard impact zones in the South of Bali, based on
a multi- scenario approach (left) and on one inundation scenario (right) (Source:
own illustration – Data basis: AWI; Map basis: Quickbird satellite image............................53
Figure 27 (left) / 28 (right) Safe areas in the study area based on historical
tsunami data (left) and on one inundation scenario (right) (Source: own
illustration – Data basis: NGDC; Map basis: Quickbird satellite image) ...............................54
Figure 29 (left) / 30 (right) Distribution of the day and night population in the
study area, calculated with Method A (Source: own illustration - Data basis:
Khomarudin et al. 2008; Map basis: Quickbird satellite image) ..........................................56
Figure 31 (left) / 32 (right) Distribution of the day and night population in the
study area, calculated with Method B (Source: own illustration - Data basis: own
data collection; Map basis: Quickbird satellite image) .......................................................56
Figure 33 (left) / 34 (right) Comparison of the population number in two
sample areas based on the calculations with Method A and B (Source: own
illustration - Data basis: own calculation; Map basis: Quickbird satellite image) ..................57
Figure 35 (left) / 36 (middle) / 37 (right) Functional urban sectors in the
study area (left); distribution of day (middle) and night (right) population in the
XIII
study area, calculated with Method B (Source: own illustration - Data basis: own
data collection; Map basis: Quickbird satellite image) ....................................................... 58
Figure 38 (left) / 39 (right) Left: Land use classes in the study area (Source:
own illustration – Data basis: LAPAN; Map basis: Quickbird satellite image; Right:
Evacuation speed per cell in second per meter (Source: own illustration – Data
basis: own calculation; Map basis: LAPAN land use classes............................................... 60
Figure 40 (left) / 41 (right) Left: Evacuation shelter accessibility based on a
CWD approach (Source: own illustration – Data basis: own calculation; Map
basis: inverse evacuation speed map); Right: Example for the definition of
“fastest evacuation routes” (Source: own illustration – Data basis: CWD surface;
Map basis: population distribution, calculated with Method B)........................................... 61
Figure 42 Criteria check list for tsunami evacuation shelter buildings (ESB)
(Source: own illustration – Data basis: BUDIARJO 2006, EISNER AND NTHMP 2001) ............... 64
Figure 43 (UL) / 44 (UR) / 45 (LL) Evacuation shelter accessibility for the day
population (calculated with Method A, B and the census population) within 15
min – (Source: own illustration - Data basis: CWD surface; Map basis: catchment
areas of the evacuation shelters, Figure 29, 31 and census population map) ..................... 65
Figure 46 (left) /47 (right) Evacuation shelter accessibility within 15 min for
day/night population, calc. with Method B (Source: own illustration - Data basis:
CWD surface; Map basis: catchment areas of evacuation shelters, Fig. 31 and 32) ............ 67
Figure 48 (left) /49 (right) Evacuation shelter accessibility within 15 min for
day/night population, calc. with Method A (Source: own illustration - Data basis:
CWD surface; Map basis: catchment areas of evacuation shelters, Fig. 29 and 30) ............ 67
Figure 50 (left) /51 (right) Evacuation ability within 15 min for the day and
night population, calculated with Method B (Source: own illustration - Data basis:
CWD surface; Map basis: catchment areas of the evacuation shelters) .............................. 69
Figure 52 (left) /53 (right) Evacuation ability within 15 min for the day and
night population, calculated with Method B (Source: own illustration - Data basis:
CWD surface; Map basis: catchment areas of the evacuation shelters) .............................. 69
Figure 54 Useful combination between hazard impact zones, warning levels and
evacuation guidance (Source: own illustration - Data basis: GTZ- IS (2008),
modified) .................................................................................................................... 77
Figure 55 Potential improvement of the CWD method (Source: own illustration -
Data basis: CWD approach)............................................................................................ 83
Figure 56 Analysis of the evacuation shelter accessibility within 15 min regarding
the population distribution (Method B) in the study area (Source: own illustration
XIV
- Data basis: CWD surface and population data calculated with Method B; Map
basis: Figure 31 and 46) .................................................................................................86
Figure 57 Analysis of the evacuation shelter accessibility within 15 min regarding
the population distribution (Method A) in the study area (Source: own illustration
- Data basis: CWD surface and population data calculated with Method A; Map
basis: Figure 29 and 48) .................................................................................................86
Figure 58 (left) / 59 (right) Left: Evacuation bottlenecks in the study area
(Source: own illustration - Data basis: population distribution calculated with
Method B and the evaluated road network; Map basis: Figure 31); Right:
Functional urban sectors in the study area (Source: own illustration - Data basis:
own data collection; Map basis: Quickbird satellite image) ................................................87
XV
List of tables
Table 1 Research objectives of the thesis (Source: own composition).............................. 2
Table 2 Research Objectives and research questions of the thesis (Source:
own composition) ............................................................................................................ 3
Table 3 Data requirement and data acquisition methods in relation to the
research questions (Source: own composition) ................................................................ 16
Table 4 Preliminary functional sectors in the study area (Source: own
composition - Data basis: Quickbird satellite image)......................................................... 17
Table 5 Main contents of the working sheet for the fieldwork (Source: own
composition).................................................................................................................. 18
Table 6 Valuation criteria for the road network in the study area (Source: own
composition - Data basis: Field work in the study area).................................................... 19
Table 7 Overview of potentially suitable facilities for vertical evacuation in the
study area (Source: own composition) ............................................................................ 21
Table 8 Overview of potentially suitable facilities for horizontal evacuation in
the study area (Source: own composition)....................................................................... 22
Table 9 Additional criteria for evacuation shelters (Source: own composition -
Data basis: Budiarjo 2006, Eisner and NTHMP 2001)........................................................ 22
Table 10 Methods for population disaggregation (Source: own composition -
Data basis: Taubenböck et al. 2007) ............................................................................... 29
Table 11 Considered data for the calculation of the tourist population in the
study area (Source: own composition) ............................................................................ 31
Table 12 Available and required data for the evacuation modelling in the study
area (Source: own composition) ..................................................................................... 34
Table 13 Estimated speed conservation values for the road network in the study
area (Source: own composition - Data basis: own estimation) .......................................... 36
Table 14 Reclassified land use classes for the definition of speed conservation
values (Source: own composition - Data basis: ADPC (2007a), modified) .......................... 37
Table 15 Conducted calculations for the shelter accessibility in the study area
(Source: own composition)............................................................................................. 41
Table 16 Potential ESB’s in the study area, identified during the field work
(Source: own composition)............................................................................................. 48
Table 17 Questioning results from the school sector (Source: own composition).............. 51
XVI
Table 18 Differences between the day and night population calculated with
Method B (Source: own composition – Data basis: own population disaggregation
concept) .....................................................................................................................59
Table 19 Time required walking through one cell (1m) of each land use class
(Source: own composition) .............................................................................................59
Table 20 Calculation results for the shelter accessibility of each catchment area
(Source: own composition – Data basis: own calculation) .................................................66
Table 21 (left) / 22 (right) Difference between day and night accessibility
values, calculated with Method A (left) and B (right) (Source: own composition –
Data basis: own calculation) .........................................................................................68
Table 23 Difference between the evacuation ability of the day and night
population, calculated with Method A (Source: own composition – Data basis:
own calculation) .............................................................................................................70
Table 24 Difference between the evacuation ability of the day and night
population, calculated with Method B (Source: own composition – Data basis:
own calculation) .............................................................................................................70
Table 25 Data requirement and additional useful data for the thesis (Source:
own composition) ...........................................................................................................71
Table 26 Average evacuation speed values from literature (Source: own
composition) ..................................................................................................................82
Table 27 Exemplarily calculation for additional required evacuation shelters in
the study area (Source: own composition – Data basis: population distribution
concept (cp. Fig. 13).......................................................................................................89
Table 28 Reference of achievements obtained in answering the research
questions (Source: own composition) ..............................................................................96
1
1 Introduction
The devastating tsunami of December 26, 2004 caused 230 000 deaths and about $ 4.5 billion
damages. This tremendous disaster initiated – besides major relief efforts – intensive tsunami
research and the apparent need for tsunami early warning systems. As a key element for tsunami
hazard mitigation, a tsunami early warning system (GITEWS) will be developed for the Indian
Ocean in the scope of a German- Indonesian cooperation. A fundamental aspect of the project is
capacity building in the area of disaster management for decision makers, experts and the general
public at risk. The administrative bodies and the affected population have to be prepared in case
of a disaster, have to understand the warnings and react appropriately.
Evacuation of people in risk areas is the first priority once a tsunami early warning is received or
natural warning signs indicate the immediate arrival of a tsunami wave. As the available time span
between a warning and the impact of a tsunami wave in Indonesia generally is very short, all
necessary preparations should have been made in advance. An official evacuation plan is essential
to provide the community with the necessary reference, guidance and information by describing
activities and measures taken to ensure temporary evacuation of people and property from
threatened locations before the disaster strikes. During a tsunami event, the primary task of the
emergency planner is to ensure the safety and rapid evacuation of people. The decision to eva-
cuate very much depends on the severity of the immediate effects of the disaster on the vul-
nerable population.
Spatial analyses in a geo information system are a widely used approach in emergency planning by
providing features to centralize and visually display spatial information. Evacuation planning should
be based on a hazard assessment by analysing possible hazard scenarios affecting a certain area
and should develop procedures for safe transfer from the affected area into safe areas. A spatial
information system will be developed for the study area providing possibilities to visualize activities
and measures taken to ensure temporary evacuation of people from threatened locations before
the disaster strikes.
Extensive spatial information is needed to provide a suitable data base for dealing with the dy-
namic and uncertain nature of an evacuation process. Static evacuation plans are not always suit-
able when an actual disaster happens. Instructions that do not consider the evolution of a tsunami
threat over time can result in suboptimal decisions that can lead to unnecessarily imposed risk and
lost lives. In addition, people’s evacuation behaviours such as evacuation time and walking speed
also have effect on the optimal strategy. Therefore, in determining the evacuation route, it is im-
portant to explicitly consider the time-varying nature of node and people’s reaction to evacuation
in such circumstances. It is necessary to provide dynamic information and instructions to help
people to make the right decision.
2
1.1 Objectives and structure of the thesis
The aim of this thesis is to implement a pre- operating study for a spatial information system for
tsunami evacuation planning in South- West Bali in cooperation with local authorities using geo
information system and earth observation analyses. The research will be conducted under the
premise of transferability to other coastal areas with similar environmental properties especially for
the coastal areas in Indonesia exposed to tsunami impacts. The research mainly focuses on the
objectives shown in Table 1 below.
No. Research objectives
1. Definition of requirements for evacuation planning
2. Development of a transferable and applicable spatial information system for tsunami evacuation planning
3. Giving a statement about the endangerment of the study area
4. Giving a decision support for evacuation in case of a tsunami event
5. Cooperation with the local administration and decision makers
Table 1 Research objectives of the thesis (Source: own composition)
The work will consider three major parts:
Part 1 is the acquisition and implementation of geobasisdata into a Geo Information System
(GIS). Beside this, a tsunami hazard assessment has to be conducted determining the exposed
areas. This includes historical tsunami impact information and tsunami inundation modelling
results. Based on these results, spatial analyses in respect to evacuation planning considering
spatial properties of the urban and rural structure and visualization of structures potentially
suitable for evacuation are presented. The existence of adequate reachable higher ground outside
inundation areas for horizontal evacuation as well as characteristics of buildings potentially suitable
for vertical evacuation are significant information for evacuation planning and therefore are
particularly featured.
Part 2 incorporates population characteristics in the analysis. The vulnerability of population in
areas under the effect of tsunami hazards is a major reason of concern. Information on
population, required for evacuation planning, has an essential spatial component and, particularly
in urban areas, is changing continuously. Therefore available population data are disaggregated
into smaller units considering day and night distribution.
Part 3 is based on the previous results. Tsunami evacuation is simulated using both accessibility
modelling, since the modelling demonstrates the effort of people to move themselves from their
initial location to the nearest (in terms of cost) evacuation shelter (horizontal and vertical), and a
modelling of the evacuation ability considering the certain shelter capacities.
3
Research questions
In order to achieve the research objectives, the following research questions shown in Table 2 will
be addressed.
No. Research Objectives Research Questions (RQ)
1. Definition of requirements for evacuation planning
a) What spatial data are needed for evacuation planning?
b) What are the most important needs for the end users (local stakeholders, exposed population)?
2. Development of a transferable and applicable spatial information system for tsunami evacuation planning
a) Which spatial information tools are suitable for evacuation modelling?
b) What has to be considered in developing a spatial information tool to ensure its transferability to other coastal areas?
c) Which spatial parameters have to be considered for tsunami evacuation modelling?
3. Giving a statement about the endangerment of the study area
a) How is the intensity of tsunami hazard locally distributed?
b) How many people are exposed?
c) In the context of evacuation planning, where are critical facilities?
d) Where are evacuation bottlenecks?
4. Giving a decision support for evacuation in case of a tsunami event
a) Where are potential evacuation shelters?
b) Where are potential evacuation routes?
c) Are the people in risk areas able to evacuate in a given period of time?
d) Which spatial information products are useful for the local decision makers?
5. Cooperation with the local administration and decision makers
a) Which stakeholders should be incorporated in the research process?
b) How to transfer the gained knowledge and create acceptance for the local population?
Table 2 Research Objectives and research questions of the thesis (Source: own composition)
4
1.2 State of research
With a view to the research objectives, two main topics take centre stage in this thesis.
(1) Population disaggregation methods and (2) evacuation modelling approaches including
evacuation shelter properties will be discussed in this chapter regarding the current state of
research.
Population disaggregation
Awareness and preparedness to natural hazards is the critical issue for local populations who are
potentially vulnerable to a tsunami hazard. Essential to evacuation planning is determining where
people are likely to be located, which varies temporally (e.g. day and night).
The census district is the common basis for collecting population data, also for the study area in
this thesis. Population data from censuses are commonly made available per administrative or
political unit whereby the populations seem to be homogeneously distributed over the whole area,
despite possibly significant variations in real population densities (SCHNEIDERBAUER 2007).
However, evacuation planning requires detailed information on spatial and temporal population
distribution, such that the idea to safe all the people in the tsunami hazard area can be achieved
(BUDIARJO 2006).
Two methods for disaggregating population data are generally available. A top- down approach
interpolates the total population on a local scale based on physical structures mapped from remote
sensing data. A bottom- up approach extrapolates punctual information onto the district level
(TAUBENBÖCK ET AL. 2007). The basic idea behind spatial disaggregation using remote sensing data
is based on a correlation between the structural characteristics of the urban or rural environment
and its population.
With spatial knowledge about the physical urban morphology, regionalization can be performed.
Regionalization or localization stands in social research for an area that is subdivided into
homogeneous sectors based on economic, structural, demographic and/or social criteria. General
rules for a spatial definition of homogeneous urban sectors are:
(1) the characterization of the land use
(2) the exclusion of other land cover types such as bare soil or water, and
(3) the classification of large units with respect to the image resolution and to boundaries
along relevant natural and anthropogenic features
The result generates functionally or structurally zoned spatial sectors, subdividing the whole area
by certain homogeneous characteristics (TAUBENBÖCK ET AL. 2007). These static physical urban
elements allow a linkage to the dynamic urban element: population. The hypothesis is based on
the assumption that populations living in areas showing nearly similar housing conditions will have
homogeneous social and demographic characteristics. Thus, spatial disaggregation is based on a
5
correlation of spatial structural characteristics and population (TAUBENBÖCK ET AL. 2007). Remote
sensing can provide criteria to generate homogeneous spatial sectors within the heterogeneous
urban environment. The basic product of remote sensing is an urban land cover classification for
identification of the housing areas and their respective spatial reference. Thus, the capabilities of
remote sensing provide physical parameters to describe regularities and irregularities for
regionalization. Value-added remote sensing products provide physical criteria like ‘built-up
density’, ‘building height’, ‘land use’ and ‘location’ to classify urban morphology. Based on the
hypothesis that homogeneous structural urban sectors show homogeneous demographic
characteristics, the local population distribution is indirectly derived.
Evacuation modelling
Traditional evacuation research has primarily focused on analyses of static evacuation plans or
maps and the use of performance- based tests to determine disaster impacts on transportation
infrastructure, property, population behaviour and survivability. The use of static evacuation plans
has not yielded a convenient means to conduct experimental designs for determining the plans’
ability to accommodate population growth, urban growth, or aging populations. In addition, static
evacuation plans do not provide the best means for optimizing emergency procedures and
evacuation personnel requirements (PITTMANN ET AL. 2006).
The potential role for GIS in evacuation research has been noted by a number of authors (GATRELL
AND VINCENT 1991, DANGERMOND 1991, JOHNSON 1992, REJESKI 1993). GIS have been applied in
generating alternative evacuation routes out of a given zone and in managing the spatial data
associated with an evacuation decision support system (DE SILVA ET AL. 1993). In general, the
wider application of GIS in hazards research has focused on modelling the physical aspects of
hazards and not on potential evacuation difficulties. Although evacuation vulnerability modelling is
clearly related to GIS natural hazards research, it’s more closely aligned with GIS research on
modelling human vulnerability and risk (BURKE 1993). However, these capabilities propose GIS
technologies as a natural framework to deal with much of the complexity embedded into the
evacuation research and decision-making processes.
Simulation models represent another category of tools, which may be used for planning or decision
support. Several GIS packages offer simulation capabilities for simple processes, but specialized
models will be required to refine the level of analysis, or may represent consolidated tools that are
pre-existing to GIS in a given context (COVA ET AL. 1997). At the heart of any network evacuation
simulator is a traffic simulation model. The approaches to the design of traffic models in general
depend on the area of their intended application and the detail required in modelling the behaviour
of the entities within the system. Network- based traffic simulation models can be categorized into
three basic groups: micro, macro and meso, based on how they attempt to model the behaviour of
evacuation entities. Micro- simulators track the movement of individuals as the simulation
proceeds while macro- simulators are driven by a model analogous with fluid dynamic flows. Meso-
6
simulators attempt to gain the best of both micro- and macro- simulation by simulating “platoons”
of individuals throughout the network (DE SILVA ET AL. 2000).
The question how to combine the enormous potential of GIS in evacuation modelling with the
technical specialized and dynamic simulation models is obvious.
BATTY (1994) discusses the possible contribution GIS can make to visualize simulation. To
extremes can be identified: (1) one where the various components of GIS can be drawn into the
modelling process and (2) where modelling can take place within the GIS.
A combination of the two is often used for the design of Spatial Decision Support Systems (SDSS)
whereby using the analytical tools within GIS, some sort of analytical modelling of the data take
place. In contrast, decision models within the SDSS are likely to draw on the display and data
analyses facilities of the GIS. It is therefore evident that the union of these technologies offers
attractive modelling and analytical resources to develop a decision aid tool for evacuation planning
(DE SILVA ET AL. 2000).
For this reason, GIS and simulation models are often proposed to the decision-maker as an
“integrated tool.” This relation is presented in Figure 1.
Figure 1 Combination of GIS and dynamic simulation models (Source: own illustration)
7
1.3 Approach
1.3.1 GIS for Natural Disaster Management
Natural disasters are inevitable and it is almost impossible to fully recoup the damage caused by
disasters. But it is possible to minimise the potential risk by developing disaster early warning
strategies, prepare and implement developmental plans to provide resilience to such disasters and
to help in rehabilitation and post disaster reduction.
The use of GIS has become an integrated, well developed and successful tool in disaster
management as it allows the combination of the different kinds of spatial data, with non- spatial
data, attribute data and use them as important information in the various stages of disaster
management. A complete strategy is required to effectively reduce the impact of natural disasters,
which can be split into six phases: Disaster prevention and disaster preparedness take place before
a disaster occur, the response phase during a disaster, the disaster relief, rehabilitation and
reconstruction after the occurrence of a disaster. In the disaster prevention phase, GIS is used to
manage the large volume of data needed for the hazard and risk assessment, in the disaster
preparedness phase it is a tool for evacuation planning and the integration of satellite data with
other relevant data in the design of a disaster warning system. During a disaster, GIS can provide
essential spatial information for evacuation coordination, e.g. by calculating suitable shelter
accessibility. In the disaster relief phase, a GIS is very useful for rapid acquisition, processing and
analyses of satellite data and the provision of satellite based information products on natural and
environmental disasters, for human relief activities. In the disaster rehabilitation phase, GIS is
used to organise the damage information (BANGER 2002).
The importance of GIS for Natural Disaster Management is therefore relevant in two major
aspects: 1) the analytical capability for decision making and 2) the data integration capacity. Both
aspects allow the integrated analyses of large amounts of different data in each disaster phase
(CASTELLANOS- ABELLA 2002).
1.3.2 Tsunami evacuation planning and spatial decision support
In case of a tsunami event, the primary task of the emergency planner is to ensure the safety and
rapid evacuation of the people at risk. The decision to evacuate very much depends on the
estimated degree of the immediate impact of the tsunami on the exposed population. Planning for
a tsunami evacuation therefore involves addressing both the ability to evacuate (age, gender, etc.)
and logistical issues that greatly influence the evacuation operation (DE SILVA ET AL. 2000).
A spatial information system in a GIS environment can provide a decision support by answering
questions like, what evacuation routes should be selected based on the location of the hazard,
how many people have to be evacuated, and where they should go to reach the nearest shelter.
Testing a variety of inundation scenarios, for example under simulated emergency conditions can
also be greatly aided by GIS data and techniques. During a tsunami event, GIS can be an
8
important aid to emergency evacuation. By overlaying and analyzing a variety of maps, emergency
planners can quickly estimate populations needing to be evacuated (COLE ET AL. 2005). If the
tsunami hazard effects new areas, alternative evacuation plans or inundation scenarios can also be
implemented. The results of the GIS- generated analyses, whether as a map or report, can be
distributed instantaneously to relevant authorities over the internet for fast public information
delivery.
9
2 Study area – Case study Kuta
2.1 Geographical position and political structuring
Bali, as the westernmost island of the Small Sunda islands, is located between 0°30’40” - 8°50’48”’
southern latitude and 114°25’53” - 115°42’40” eastern longitude with an island expanse of 5 561
km². The largest east- west expansion amounts 144 km and that in north- south direction 87 km.
In the West, Bali is separated from the Java island by the small Selat Bali (Bali- street) and in the
East the 40 km wide Selat Lombok (Lombok- Street) is the parting line to the neighbour island
Lombok. The island is washed round by the Bali Sea in the North and the Indian Ocean in the
South (BAKOSURTANAL 2003).
The province of Bali is subdivided into nine administrative districts (Kabupaten), 51 sub- districts
(Kecamatan) and 570 villages (Desa/ Kelurahan) and urban districts. The villages are subdivided
again into 1 480 smaller villages (Desa adat) and about 3 630 village districts (Banjar). The
smallest official administrative unit is the village with a mayor at the top. However, the village
districts have a great importance as a mutually supportive group and local self- administration. The
capital and administration centre of the island is Denpasar with a current population of 340 000
people. Since 1992, the city is the ninth administrative district of Bali (NADLER 2005). The study
area is located in the South of the island between 8°42’36” – 8°44’29” southern latitude and
115°9’38” – 115°10’48” eastern longitude and belongs to the district of Badung. The area contains
the villages of Kuta and Legian as well as parts of Seminyak and Tuban. The common superior
administration unit is the sub- district (Kecamatan) of Kuta.
Figure 2 Geographical position of the study area (Source: own illustration - Data basis: BPS CENSUS 2000; Map basis: Quickbird satellite image)
10
2.2 Economic factor tourism
Bali is characterized by agriculture and tourism. About 50% of the population are working in the
tourism industry, the rest almost without exception in the agricultural sector. 65% of the island
area is used agriculturally, but the share of the agriculture in the Balinese gross national product
sank from about 55% in 1979 to 30% in 2005. This is mainly a result of the high growth in
population and changed land claims. In the last decades, tourism has emerged to the most
important economic factor. Considering other industries which directly depend on the touristic
demand like souvenir trade, tourist guides and transport industry, about one-third of the gross
national product is obtained directly from tourism. As a result of the multiplier factors of the tourist
expenses, growth impulses overlap to other economic sectors. The agriculture and fishery sector
cover a high proportion of the food demand from the hotels. The local beverage industry is an
important hotel supplier and the local building material production benefit from the fact that many
hotels are built in a Balinese style and with local building materials.
Figure 3 Bali – Main source of income of most of the population (Source: own illustration - Data basis: BPS PODES 2005; Map basis: BPS census 2000)
Nearly 20% of all touristic demands in the last years felt upon the souvenir purchase. The
production and the sale of local handicraft is vitally important, because the products belong to the
most preferred articles of tourists. The auxiliary incomes from these jobs show a wide social and
spatial spread, also villages apart from the tourist centres benefit from this business. Many villages
11
live on this sector and many crofters depend on this auxiliary income, because of the decreasing
income from the agriculture. However, Bali shows distinctive spatial disparities. The differences
between the districts regarding the per capita gross national product are substantially. The tourism
industry expanded in the economical advantaged and cultural rich areas in the South, whereas the
periphery was little attended. Despite a bigger spatial spread of tourist attractions in the last
decade, the touristic demand is still very concentrated to South- Bali. About 80% of the
employees in the hospitality industry are working in the district of Badung and the “tourist
triangle” Kuta- Sanur- Nusa Dua by now is mainly a complete urbanized area that merges
seamlessly in the urban agglomeration of the capital Denpasar. Therefore from one perspective
the tourism sector contributes to the reduction of regional and social disparities in Bali but
considered differently the sectorial and spatial concentration in the South will further increase
because of the economic and cultural supremacy (VORLAUFER 1999).
2.3 Tectonic structures
The Indonesian archipelago is located along a complex convergence of the Eurasian, India-
Australian, Caroline and Philippine sea plate and several minor ones. These plates are moving
relative to each other in a complicated manner. Typical plate motion can be summarized as
follows: The relative motion of the Eurasian plate to the India- Australian one is approximately
northward, whereas that of the Philippine sea plate is west by northwest. The Philippine sea plate
relative to the Caroline one is moving east by southeast near the Aru trench and west by
northwest near the Yap trench. The Indonesian archipelago consists of five active island arcs: The
Sunda, Banda, Sangihe, Halmahera and North Sulawesi arc. The Sunda arc, as a result of the
convergence of the Indian Ocean and the Eurasian plate, extends westward from Sumba through
Java, Sumatra and the Andaman islands. The Banda arc resulted from the collision of the south-
eastern part of the Eurasian and Australian plate and extends eastward from Sumba. To the
northern part of this arc, the Sangihe and Halmahera arcs in the Molluca Sea region were caused
by the activities of two opposing subductions of the Molluca sea plate. Taking into consideration
the above tectonics, the Indonesian region is divisible into 6 zones (LATIEF ET AL. 2000):
Zone A: The West Sunda arc includes the northwest Sunda Strait (Sumatra and the
Andaman Islands)
Zone B: The East Sunda arc includes the area in the region of the East Sunda Strait to
Sumba (Java, Bali, Lombok, Sumbawa and Sumba)
Zone C: The Banda arc covers the area of the Banda sea (Flores, the Timor and Banda
islands, the Tanimbar islands, Ceram and Buru)
Zone D: The Makassar Strait
Zone E: The Molucca sea, Sangihe and Halmaher
Zone F: The North Irian Jaya
12
Figure 4 Seismotectonics of 6 Indonesian zones (A- F) (Source: LATIEF ET AL. (2000), p.28)
2.4 Tsunami generation and zonal activities in the region of Bali
Tsunamis are generated by several kinds of geo-
physical phenomena; earthquakes, volcanoes,
landslides and debris flows. In Indonesia, earth-
quakes are the primary source of tsunami
generation. When the earthquake occurs, it causes
a vertical movement of the sea floor over a large
area. As the land mass shifts, the waves become
broader and move outwards. The most destructive
tsunamis are generated by large, shallow earth-
quakes that have an epicentre near or on the
ocean floor usually along subduction boundaries
where tectonic plates collide in regions of high
seismic activity. The amount of vertical and hori-
zontal motion of the sea floor and the efficiency
with which energy is transferred from the earth’s
crust to the ocean water to create destructive
waves, are all part of the tsunami generation
mechanism. In deep water, a tsunami travels over
great distances at high speed and with limited
energy loss. As the wave approaches land, its
height increases while its speed decreases.
Figure 5 Tsunami generation and propagation (Source: http://www.important.ca/tsunami_causes.html)
13
As a result of the complicated plate- convergence presented in Chapter 2.3, the Indonesian region
has a very high seismic activity and will continue to experience big earthquakes in the centuries to
come, some of which could generate tsunamis.
Bali is located in the area of the East Sunda arc (see Zone B in Chapter 2.3). In the following, the
occurrence of historical tsunamis caused by earthquakes and affecting Bali is presented. The seis-
micity in the East Sunda arc is mainly controlled by the action of the Indian Ocean plate sub-
ducting beneath the Eurasian plate. The seismicity depth reaches about 650 km with a gap in
seismicity between 300 to 500 km. 82 destructive earthquakes which have the potential to gene-
rate a tsunami have been reported in this zone, about 45% of the total number of destructive
earthquakes reported in Indonesia. Figure 6 shows the tsunami events in the East Sunda arc from
1600 to 1998. About 10% of the total number of tsunamis reported in Indonesia occurred in this
zone. Nine tsunamis were generated by earthquakes and one by volcanic eruption. The last
tsunami was the East Java tsunami in 1994 (Moment Magnitude M = 7.6) that killed 238 people.
The average interval between tsunami events is estimated to be about 10-15 years (LATIEF ET AL.
2000).
Figure 6 Histogram of destructive earthquakes (1800 – 1999) and tsunamis caused by earthquakes (1600 – 1999) by zone and by decade (Source: LATIEF ET AL. (2000), p. 33, modified)
15
3 Methods
3.1 Data collection
The data research was conducted during a five month stay in Bali. A cooperation with GTZ-IS
Indonesia, a German stakeholder in the course of the GITEWS project, was agreed in advance. In
the context of capacity building, GTZ- IS is responsible for the strengthening of the local organi-
sation related to tsunami warning and disaster management in Indonesia. The close collaboration
on site was important for the first contacting with local stakeholders and the official legitimation of
the research process.
3.1.1 Data requirement and data acquisition methods
The data collection process was conducted before and during the fieldwork period. Before the field
work, the data were obtained from literatures and internet sources. During the fieldwork period,
data acquisition methods included interviews and field observation as well as data acquisition for
secondary sources. Table 3 provides the list of data requirement and data acquisition methods in
relation to the research questions.
No. Research Question (RQ)
Data requirement Source Acquisition method
1. RQ 1a, 1b
Required spatial information for an evacuation planning process
- Reference on tsunami evacuation
- Reference on existing institutions with experience in dealing with tsunami evacuation planning
- FEMA/NOAA –IOC, UN-ISDR
- PMI
- BUDIARJO (2006)
- Herryal Z. Anwar (LIPI)
- Literature study
- Internet browsing
- Expert interviews
2. RQ 2a, 2b, 2c
Spatial information systems for evacuation planning
- Information about evacuation modelling using GIS
- Reference on institutions dealing with spatial information and GIS
- Availability of suitable software products
- ADPC (2007), (2007a)
- BUDIARJO (2006)
- JONG ET AL. (1996)
- Oak Ridge National Laboratory
- BAPPEDA
- Literature study
- Internet browsing
- Expert interviews
16
3. RQ 3a, 3b, 3c, 3d
Components of risk assessment
- Tsunami hazard parameter
- Information about population distribution
- Critical facilities locations
- DLR
- LIPI
- Field observation
- H.Spahn (GTZ)
- Khomarudin et al. (2008)
- Internet browsing
- Expert interviews
- Facility identification
- Functional urban zone identification
4. RQ 4a, 4b, 4c, 4d
Evacuation modelling
- Evacuation routes
- Evacuation shelter locations
- Evacuation shelter capacity
- BAKOSURTANAL
- Field observation
- AWI
- POST ET AL. (2007)
- BUDIARJO (2006)
- Internet browsing
- Expert interviews
- Road and facility identification
5. RQ 5a, 5b
Cooperation process in the context of capacity building
- Reference on local stakeholders
- Results of cooperation process
- Kelurahan Kuta
- BMG
- PMI
- GTZ
- IHRA
- KESBANGLINMAS Kabupaten Badung
- Expert interviews
- Workshop participation
- Internet browsing
Table 3 Data requirement and data acquisition methods in relation to the research questions (Source: own composition)
3.1.2 Data acquisition and field work preparation
Before the fieldwork, data collection mainly focused on the secondary data, collecting information
about existing knowledge on tsunami evacuation, socio- economic, socio- demographic and spatial
data of the study area and the touristic infrastructure of whole Bali. Based on these acquired data,
an initial analysis was conducted to prepare the data acquisition in the field. The preparation
included:
Developing interview guidelines
Referring to the RQ – block 3 in Table 3, the locations of critical facilities and potential safe areas
in the context of evacuation planning had to be determined and valuated concerning their critical
characteristics. As critical facilities, schools and hospitals were considered because of their strong
concentration of people at certain times during the day.
Due to the dominating touristic infrastructure in the study area, hotels are of particular importance
for vertical evacuation. Their locations could be determined by the information from statistical
data, but specific characteristics had to be gained by interviews. Necessary information and
knowledge to obtain were formulated in an interview guideline and translated into interview
questions.
17
Preparation for workshops on evacuation planning
Referring to the RQ- block 1, cooperation processes with local administration and local decision
makers in the form of technical workshops on evacuation planning were very important for the
exchange of experiences and information. Main components of the field work process, especially
methods for the disaggregation of population data and the identification of potential evacuation
shelters, had to be discussed. A key topic was the implementation of technical evacuation
components in a general evacuation planning concept as well as the clear and understandable
presentation of evacuation procedures which have to be disseminated to the local population.
Functional zonation of the study area
Referring to the RQ –block 3, detailed information about the population distribution in the study
area had to be gained by fieldwork. An own population distribution concept, considering
information about population movements between different functional urban sectors was
developed to disaggregate population data gained from census data on village level. This concept
is explained more detailed in Chapter 3.3. On the basis of a Quickbird satellite image, different
functional urban sectors were signed out in the study area to identify homogeneous zones with
similar urban characteristics (cp. Chapter 1.3). The considered parameters for this zonal
predefinition were building density, building type, building size and roof colour. However, due to a
lack of time and technical expertises, this identification process is not based on a methodological
but only on a visual approach.
In this preparing step, five different sectorial uses could be determined and signed out
accordingly:
Tourism sector Local sector
Commercial area Commercial area
Residential area Residential area
Mixed tourism and local sector
Table 4 Preliminary functional sectors in the study area (Source: own composition - Data basis: Quickbird satellite image)
The pre-zonation was cross- checked and corrected during the fieldwork process. Relevant
secondary input data could be gained from the “Central Bureau of Statistics Indonesia (BPS)” and
the mayor of the village of Kuta.
18
Preparing working sheets for the fieldwork
Referring to the RQ- block 3 and 4, a comprehensive fieldwork had to be conducted. To ensure a
consistent research, a detailed working sheet was developed. During the preparation, useful data
from the “Indonesian Red Cross (PMI)”, which already conducted similar fieldworks, could be used
as helpful examples and proposals for the detailed development of the sheet. The main contents
are stated below.
A Functional Zoning Verification of predefined sectors
Subzonation if necessary
B Road network Numbering of road segments
Number of lanes
Traffic volume
C Potential safe areas for vertical evacuation Building function
Number of storeys
Potential capacity
D Potential safe areas for horizontal evacuation Type of location
Potential capacity
E Critical facilities Facilities name
Facilities function
Number of storeys
Potential capacity
Table 5 Main contents of the working sheet for the fieldwork (Source: own composition)
3.1.3 Field work
The field work was conducted in the study area of the Kecamatan (Sub- district) Kuta. The main
objective of the survey was the verification of the predefined functional urban sectors, the
identification of the road network, potential safe areas and critical facilities in the sense of tsunami
evacuation.
Identification of the road network
For the identification of the road network, detailed road maps as well as street layers, digitized
from satellite images, were already available. The idea of this activity was to complete the
available road dataset and to valuate certain road segments regarding their important function as
evacuation route. Therefore all roads in the study area were evaluated considering their width,
which is described with the number of lanes, and the traffic volume during day and night. The
evaluation criteria and the considered attributes are presented in Table 6.
19
Road Network Width ID Traffic volume ID One lane small 1 Very much 1 One lane wide 2 Much 2 Two lanes small 3 Not much 3 Two lanes wide 4 Four lanes 5
Table 6 Valuation criteria for the road network in the study area (Source: own composition - Data basis: Field work in the study area)
Identification of functional urban sectors
Information on population, required for evacuation planning, has an essential spatial component
and, particularly in urban areas, is changing continuously. The census data on village level are the
most detailed available information, but for evacuation planning at local level these data are too
coarse to provide a good insight into the spatial distribution of the population at risk. The sub-
division of the study area into functional urban sectors was necessary to get a more detailed
impression of the population movement within one day (HOFSTEE AND ISLAM 2004). As explained in
Chapter 3.1.2., the predefined sectors in the satellite image had to be verified in the fieldwork. For
the most part the predefined sectors had to be subdivided again. Unclear referable sectors were
defined by interviewing local people on site. The sector identification process is shown in Figure 7.
Figure 7 Identification of functional urban sectors in the study area, gained by field work (Source: own illustration)
20
Identification of evacuation shelters
a) Building functions potentially suitable as evacuation shelter building (ESB)
The observation aimed at identifying building and location functions that can be assigned as
evacuation shelter. In case of a tsunami event, two methods to evacuate people from the hazard
area are generally available (EISNER AND NTHMP 2001):
1. Horizontal evacuation - moving people to more distant locations or higher ground outside the
inundated area
2. Vertical evacuation - moving people to higher floors in buildings.
Referring to general characteristics of suitable shelters gained from literature (BUDIARJO 2006,
EISNER AND NTHMP 2001), the criteria of observation are:
=> for horizontal evacuation: public locations with an adequate capacity
=> for vertical evacuation: public facilities, multi-storey buildings, available reserve space for
temporary evacuation and good construction quality.
However, these general characteristics for evacuation shelters which are considered for the field
work indeed are correct, but a professional technical expertise is necessary for an official definition
of a building as evacuation shelter. An overview of facilities and locations in the study area which
meet these basic demands for evacuation shelters is presented in Table 7 and Table 8.
No. Building function
Suitability for vertical evacuation
Public- oriented function
Building design and construction
Critical issue
1. Hotel Suitable Limited access to public
- Hall, foyer and function room can be occupied for evacuation
- Generally well planned and good construction; has to be proved!
Security and privacy issues should be arranged together with owner and disaster authority
2. School Suitable Accommodate education activity for students living in surrounding area
- Hall and classes can be occupied for evacuation
- Generally well-planned and good construction; has to be proved!
During teaching time the building can only accommodate the people in it
3. Government office
Suitable Civil building, oriented to serve the people
- Hall, foyer and function room can be occupied for evacuation
21
4. Sport hall Suitable Accommodate sport activities of the users
- Open lay-out suitable for accommodating huge numbers of evacuees
- Generally well-planned and good construction; has to be proved!
5. Market building
Suitable only for people within the building
Public facilities, accessible to everybody in the certain operating hours
- Commercial oriented, lack of empty space, full of merchandise and storage
During evacuation, the building can only accommodate the people in it
6. Mosque Suitable Accommodate prayer activities for Moslems
- Open lay-out suitable for accommodating huge numbers of evacuees
- Generally well planed and good construction; has to be proved!
7. Shopping centre
Suitable only for people within the building
Public facilities, accessible to everybody in a certain operating hours
- Commercial oriented, lack of empty space, full of merchandise and storage
- Generally well planned and good construction; has to be proved!
Vulnerable to robbery and thievery in emergency situation. During evacuation, the building can only accommodate the people in it
8. Billiard centre
Suitable Public facilities, accessible to everybody in the certain operating hours
- Open lay-out suitable for accommodating huge numbers of evacuees
Table 7 Overview of potentially suitable facilities for vertical evacuation in the study area (Source: own composition)
22
Table 8 Overview of potentially suitable facilities for horizontal evacuation in the study area (Source: own composition)
b) Potential evacuation shelter identification
After identifying general characteristics of suitable shelters, the criteria used to determine which
building or location is a potential evacuation shelter, have to be completed. Referring to expert
interviews and technical literature (BUDIARJO 2006, EISNER AND NTHMP 2001), the additional criteria
of observation are:
for safe areas (horizontal evacuation) for evacuation shelter buildings (ESB) (vertical evacuation)
- outside maximum inundation zone
- good accessibility
- Slope < 20° (cp. Chapter 3.2.2)
- suitable land use classes
- good building accessibility
- good accessibility to space for temporary evacuation inside the building
- located at a distance of more than 200m from the shore (BUDIARJO 2006, P.61)
- good and proved stability
Table 9 Additional criteria for evacuation shelters (Source: own composition - Data basis: Budiarjo 2006, Eisner and NTHMP 2001)
During the field work, suitable evacuation shelters were identified and valuated considering the
mentioned criteria (see Chapter 4.1). But also in this case, a professional technical expertise is
required to assign evacuation shelter buildings.
No. Location function
Suitability for horizontal evacuation
Public- oriented function Location property
Critical issue
1. Parking place Suitable Public location, accessible to everybody at any time
- Flat area
- Asphaltic ground
2. Sports field Suitable Public location, accessible to everybody at any time
- Flat area
- Flat gras vegetation
3. Public places Suitable Public location, accessible to everybody at any time
- Flat area
- Usually asphaltic ground
4. Field/ Acre Suitable Private property but mostly accessible to everybody at any time
- Rough area
23
c) Hotel survey
Due to the fact that the hotel sector in the study area is part of the private commercial sector,
there is no responsible public administration which can provide detail building and socio- economic
information. To identify a hotel as a potential evacuation shelter, a lot of data apart from the
location and general accessibility of the building are necessary. A detailed building inspection, like
for the public facilities, was not possible due to a lack of time. Therefore the “Indonesia Hotel and
Restaurant Association Badung Regency of Bali”, an umbrella organisation of star hotels in Bali,
was contacted to conduct a written survey for the topic “Hotels as evacuation shelter buildings”.
The questionnaire was developed together and is given in Appendix 1.
Identification of critical facilities
In case of a tsunami evacuation, special attention should be paid to the more vulnerable people
and groups. Particularly the very young ones as well as old and handicapped people are among
the victims of a tsunami disaster (BIRKMANN ET AL. 2007). These groups are not separately
mentioned in statistical data in the required spatial resolution, therefore statements about their
spatial distribution can only be made in relation to some known facilities. Schools and hospitals are
important institutions that house people who are not able to be responsible for their own
evacuation. As described in Chapter 3.1.2., a questionnaire for the fieldwork was developed in
advance to evaluate the facilities concerning their critical characteristics. During the fieldwork,
responsible authorities were contacted in order to get information about the geographical position
and relevant evacuation properties of the facilities. Therefore an own questionnaire only for
hospitals and schools were developed as presented in Appendix 2 and 3.
3.2 Hazard impact & exposed area
3.2.1 Tsunami hazard assessment
An assessment of the tsunami hazard is needed to identify people and assets at risk, and the level
of that risk. This assessment requires knowledge of probable tsunami sources (such as
earthquakes, landslides, volcanic eruption), their probability of occurrence, and the characteristics
of tsunamis from those sources at different places along the coast. Data of earlier (historical and
paleotsunamis) tsunamis may help quantifying these factors. In case that only very limited or no
past data exist, numerical models of tsunami inundation can provide estimates of areas that will be
flooded in the event of a local or distant tsunamigenic earthquake (UNESCO- IOC 2006).
An evacuation resulting from a tsunami hazard impact forces immediate action because of little
warning and limited preparation time. Detailed knowledge for early warning (expected arrival time
and extend of tsunami), the potential tsunami impact at coastline (expected wave heights at
coastline) as well as the potential tsunami impact on land (expected intensities on land, hazard
zones) is required for an applicable evacuation planning (POST ET AL. 2007)
24
Four methods are generally available to conduct a hazard assessment.
(1) Based on historical records of tsunami impact observations
(2) Based on one numerical inundation scenario
(3) Based on many numerical inundation scenarios
(4) Based on a combination of historical data and modelling results
For this research one tsunami inundation scenario is the basis for the evacuation modelling.
However, an assessment from historical data was conducted in advance due to a lack of
inundation scenarios at the beginning of the research. A multi- scenario approach was also taken
into consideration as a basis for further analysis. All three approaches are shortly presented in the
following.
I. Using historical records of tsunami impact observation
The National Geophysical Data Centre (NGDC) provides a global tsunami database with
information on tsunami source events and locations where effects from tsunamis were observed.
These data include the maximum water heights above sea level (in meters) and the maximum
horizontal distance of inland flooding (in meters). Based on these information and knowledge
about protective measures in the affected area, the degree of hazard impact on land can be
quantified as presented in Figure 8 below (POST ET AL. 2007).
Figure 8 Decision tree to approximate the degree of hazard impact on land (Source: Post et al. (2007), modified)
25
It is quite trivial that there is a correlation between the distance to the coast and the degree of
impact. The NGDC database can be used for a classification into distance zones with different
probabilities to be affected by a tsunami. From the results one can derive zones with a different
frequency of an event.
The influence of the altitude is obviously also an important factor. The spatial distribution of the
altitude (topography) along the coast can be derived from Digital Surface Models (DSM’s) or Digital
Terrain Models (DTM’s). This classification is strongly depending on the resolution of the surface or
terrain model.
The protective measures include all aspects (indicators) which can mitigate the tsunami intensity.
These can be structural measures like dams, but also the vegetation can have a protective effect,
such as mangrove forests. To quantify the relevance of protective measures it is necessary to
define their relative qualitative differences. This accounts especially for the vegetation, where the
spatial differences in the area of interest need to be incorporated as well as the assessment scale
of the risk assessment. For example, the existence of many or highly efficient protective measures
in a given area can lead to the estimation of “no risk” in the decision tree for a given scenario. By
the combination of these gained parameters, the degree of a tsunami impact on land can be
represent spatially distributed. (POST ET AL. 2007)
II. Using inundation modelling results
For estimating tsunami impacts on land, the hazard assessment is often conducted by using
numerical inundation modelling results. This analysis requires high resolution topography and
bathymetry data (elevation data on land and underwater), and information about potential
regional tsunami sources. The outcomes of tsunami inundation modelling are to identify areas that
could be flooded and to estimate water depths, current strengths, wave heights, and wave arrival
times. The calculated results yield the spatial distribution of the maximum inundation area
depending on the location of the source (STRUNZ ET AL. 2008A).
Based on one scenario
In order to identify the potential tsunami impact of a “worst case” event for a certain region, a
scenario with an epicentre relatively near the coast and a high moment magnitude can be used to
derive hazard impact zones expected.
This research was conducted using an inundation scenario calculated by the Alfred Wegener
Institute (AWI) for Polar and Marine Research within the GITEWS tsunami early warning project.
The underlying tsunami simulation system is based on the current knowledge on the possible
ocean bottom deformations due to seismic activity in the Sunda Trench region.
However, the decision to take only one scenario was not made in order to show a worst case
scenario, but because the small study area is completely affected by a majority of the available
scenarios. For the evacuation modelling an artificial scenario was selected showing both affected
26
and safe areas. The scenario is based on the assumption of a tsunamigenic earthquake in the
South- West of Bali with a moment magnitude of 8.5. Inundation heights in the study area lower
than 0.5m are not considered, because a life- threatening situation usually not emerges at this
level. The estimated tsunami arrival time at coastline is about 30minutes.
The used scenario is visualized in Figure 9 and Figure10.
Figure 9 (left) / 10 (right) Inundation scenario used for the evacuation modelling in the study area (Source: own illustration - Data basis: Alfred Wegener Institute (AWI), modified; Map basis: Indonesia overview, DLR)
Based on many scenarios
A multi- scenario approach analyses the likelihood of tsunamis of various sizes (based on many
scenarios) that can then be simplified into tsunami hazard zones.
An efficient way to establish tsunami hazard zones using multiple scenario results is by counting
on each point on the land if this location is inundated or not. Incorporating all available scenarios -
which should be representative for all possible tsunami events in the considered region - leads to a
frequency distribution of each location on land being exposed to a tsunami or not. Although not
directly indicating an expected intensity, the resulting information may serve decision makers to
characterise zones by the fact of frequency being exposed:
(1) High impact zone, which is hit by every potential tsunami
(2) Medium impact zone, which is hit by many potential tsunamis
(3) Low impact zone , which is only hit by very few, extreme tsunami intensities
27
Tsunami hazard zones can also be defined by relating the impact areas to tsunami warning
categories. Tsunami Early Warning Systems currently relate the level of warning (e.g. minor, major
warning) to the expected wave height at the coastline. By deriving the inundation zone for each
warning level (wave height), one can relate the warning levels to hazard impact zones and safe
areas within the hazard map (STRUNZ ET AL. 2008A).
3.2.2 Degree of exposure and potential safe areas
Based on the above presented hazard assessment results, areas which are not affected by a
tsunami wave and hence outside the inundation zone, can be regarded as potential safe areas.
Figure 11 Requirements for safe areas (horizontal evacuation) (Source: own illustration - Data basis: expert interview with Dr. Joachim Post, 15.04.2008)
Considering evacuation planning aspects, these not-affected zones have to meet further conditions
to be classified as safe area. To ensure a good accessibility and a sufficient capacity for a huge
amount of evacuees, the zones have to be larger than 10 000m² and a slope around the areas
28
smaller than 20° (EXPERT INTERVIEW DR. JOACHIM POST, 15.04.2008). An overview of the
requirements for safe areas is presented in Figure 11 considering hazard assessment components
both from historical records of tsunami impact observations and from numerical inundation
modelling.
3.3 Population modelling
As presented in Chapter 3.1.3., for evacuation planning it is essential to have a good insight into
the spatial distribution of the population at risk. The available census data on village level are too
coarse for a planning process at local level. The question then is: how can population data in
census districts be disaggregated to smaller geographical or mapping sectors, that better satisfy
the demands for information for emergency actions and disaster mitigation management. (HOFSTEE
AND ISLAM 2004)
It is quite difficult to localize population more precisely by disaggregating census data and only to
be undertaken when adequate information of the location of dwellers are available (from high-
resolution satellite images, land use maps, etc.) as there can be a high variation in area, number
and characteristics of the population. A Quickbird satellite image was available for the study area
and therefore a concept of spatial disaggregation using remote sensing data was initially applied.
An object- oriented urban land cover classification can serve as a basis to structure the complex
urban environment in homogeneous zones.
Based on two different ancillary input data – total population for the study area and punctual
population data collected from fieldwork – two methods for population disaggregation are
generally applicable. A top-down approach interpolates the total population on a local scale based
on physical urban structures mapped from remote sensing data. A bottom-up approach
extrapolates the punctual information (houses) onto the district level.
Due to a lack of time, the latter method was not practicable and therefore only a top- down
approach was reasonable. The focus was to establish a correlation between the spatial physical
characteristics of an urban area and the total population of the study area. The first step of this
approach is the spatial quantification of homogeneous units based on physical criteria to classify
urban morphology. However, to identify these elements in a GIS, a comprehensive preliminary
work is required, while an object- oriented classification methodology of built- up areas for the
detection of house units have to be conducted in advance (TAUBENBÖCK ET AL. 2007). This effort
was not applicable for the research and therefore only a visual interpretation of some physical
criteria in the satellite image was useful, as presented in Point II in this Chapter.
The common physical parameters which can be derived by an object- oriented classification
methodology are presented in Table 10 below considering the application and use for this
research.
29
Method Goal Application Use
Built- up density
Calculation and classification of the proportion of the built- up density to a defined area
Giving a statement about population density depending on built- up density
- Building mask for the study area not available
- The study area is a mixed touristic area, population density diverge widely
Visual interpretation on the basis of a QUICKBIRD image
Building size
Building classification regarding building size
Giving a statement about building use and respectively about population distribution
- Building mask for the study area not available
- Building height not available
Visual interpretation on the basis of a QUICKBIRD image
Building shape and roof structure
Building classification regarding building shape and roof structure
Deriving information about the building function and respectively about population distribution
- Very time- consuming
- The validation of classifications for other areas showed only moderate results
Visual interpretation on the basis of a QUICKBIRD image
Land use
Proportion of settlement areas inside a defined area
Deriving more detailed information about population distribution as from census data
- The study area is nearly complete settlement area and to small for useful information gained through land use data
Visual interpretation on the basis of a QUICKBIRD image
Table 10 Methods for population disaggregation (Source: own composition - Data basis: Taubenböck et al. 2007)
For the study area, two alternative disaggregation methods using a top- down approach were
tested:
I. METHOD A: Combination of land use and census data
The population distribution can be modelled combining a land use dataset with a scale 1 : 50 000
and BPS (Central Bureau of Statistics Indonesia) census data from 2000 providing population
figures on the level of the administrative unit desa/ kelurahan (village). Weighting factors
concerning the presence of people in areas of different types of land use are the critical value in
this analysis. The factors were derived from BPS PODES data (KHOMARUDIN ET AL. 2008). The
percentages of weighting indicate the number of people who perform an activity in the certain
land use classes. Therefore the weighting between rural, urban, coastal and non- coastal areas is
different, because the people activity in these areas is different. The main sources of income
indicate the percentage of people performing activities in certain regions and on the basis of these
information the different weighting factors were developed. The method is visualized in Figure 12.
30
Figure 12 Population disaggregation with a combination of land use and census data (Source: own illustration - Data basis: BPS CENSUS 2000 and LAPAN; Map basis: BPS CENSUS 2000)
II. METHOD B: Combination of remote sensing data and field work results
Another method to derive more detailed information about the population distribution is to
consider population movements between functional urban sectors during day- and night-time. As
shown in Chapter 3.1.3., different homogeneous sectors were assigned on the basis of a Quickbird
image and cross- checked during the field work regarding their main function inside the settlement
area. To quantify the population movements between these sectors, an own distribution concept
was developed. Therefore the people’s activities in the defined zones during day- and night-time
were of particular interest.
BPS census data from 2000 providing population figures on village level are the basis of this
concept. To calculate the actual population, the touristic population had to be quantified in
addition. Due to the fact that the study area is an intersection of different administrative units (cp.
Chapter 2.1), the touristic population had to be estimated considering several data sources as
presented in Table 11 below.
The actual population has to be separated in a day and night population. While the night
population conform to this calculation, for the day population the daily commuters have to be
considered. Due to the lack of dependable data for these people movements, figures were
estimated on the basis of information gained by interviews. The distribution and quantification of
the day and night population to the certain functional sectors was the result of field observations
and experts interviews with I Gede Suparta (Major of Kuta village) and Rokhis Khomarudin (LAPAN).
31
DATA SOURCE
Average of guests per day in accommodation establishments in Bali 2005
Ministry of Culture and Tourism, Indonesia
Room occupancy rate in classified hotels in Bali 2004 Ministry of Culture and Tourism, Indonesia
Bed occupancy rate in classified hotels in Bali 2003-2004 Ministry of Culture and Tourism, Indonesia
Average room occupancy in hotels in Bali from January 2001 – September 2007
Bali Tourism Development Corporation
Direct tourist arrivals to Bali from 2002 – September 2007 Bali Tourism Development Corporation
Number of foreign visitors in Kuta village 2007 Police department, Sector Kuta
Bali accommodation directory 2006 Bali Government Tourism Office
Table 11 Considered data for the calculation of the tourist population in the study area (Source: own composition)
Figure 13 shows the estimated population distribution in the study area during day- and night-time
considering the cross- checked functional zonation.
32
Figure 13 Population distribution concept for the study area considering functional urban sectors (Source: own illustration - Data basis: own data collection)
33
The effect of the different reference parameters of both methods is presented in Figure 14 below.
Crucial for the population modelling result of Method A is the strong representation of the census
population and the neglect of the tourist population. Method B is based on a much more detailed
approach and the gained results differ widely from the census data by considering tourist and
commuter data. The calculated results of both methods are presented in Chapter 4.3.
Figure 14 Effect of different reference parameters for the population modelling results in the study area (Source: own illustration)
3.4 Accessibility modelling
Accessibility can be defined as the ability for interaction or contact with sites of economic and
social opportunity (FARROW AND NELSON 2001). However, there are a multitude of ways in which
this intuitive concept has been expressed in literature. MOSLEY (1979) conceptualised the basic
components of accessibility into: (1) the people, (2) the activities or services which people require,
and (3) the transport that links between those two. The actual accessibility depends on each of
the components of the following scheme.
People Transport Activities
34
In this scheme, people in the figure represent the population. The size and the composition of the
population affect the accessibility, because they determine the scale of the demand for services or
activities. The transport links reflect the travel time, costs and effort to travel between origin and
destination location. The activities reflect the spatial distribution of activities at destinations and
the demand for those activities. At Moseley’s scheme, accessibility analyses can be applied to both
ends. JONG AND ECK (1996) rewrote this schema in Geographic Information System (GIS) terms as
follows.
Here, which origins located within destination’s reach can be calculated, and vice versa, depending
on the specific demand of the user.
In this research, accessibility modelling is the preparing process for evacuation modelling by
providing the movement surface for the evacuees. Referring to point 4 of the research objectives
(cp. Table 2), the following questions have to be answered to give a decision support for
evacuation in case of a tsunami event:
a) Where are potential evacuation shelters?
b) Where are potential evacuation routes?
c) Are the people in risk areas able to evacuate in a given period of time?
An overview of available and required data for evacuation modelling is presented below in
Table 12.
Research Question
Available data Required information
a) Potential evacuation shelters gained from historical data, inundation modelling and field work
No additional data required
b) Road network Other passable paths (off-road)
c) Population distribution calculated with two different methods
Ability to evacuate
Table 12 Available and required data for the evacuation modelling in the study area (Source: own composition)
People Transport Activities
Origin Transport Destination
Location Network Location
35
Research question a) and b) will be answered in this chapter, question c) in the subsequent
Chapter 3.5 (“Evacuation modelling”)
The following Chapter 3.4.1 describes a simple but flexible GIS method and tool for deriving
verifiable accessibility models in the context of evacuation modelling.
3.4.1 Concept of Cost Weighted Distance (CWD)
To define the best evacuation route from a given point, the fastest path from that point to the
arrival point has to be found. The fastest path is not always the shortest path (which is the direct
line between point and the arrival point).
In this concept the accessibility is calculated on a cost surface which consists of a regular two-
dimensional grid where each cell represents either passable routes such us roads, vegetation and
beach or relatively inaccessible land and water bodies. Different types of evacuation routes have
different characteristics. A flat road, for example, allows faster travel speed than dense vegetation.
Therefore it is not enough to measure the distance between two points. Instead, a measure of
travel cost is preferable, which can be considered as travel time (ADPC 2007A).
Figure 15 Computation of CWD between two adjacent cells (Source: ADPC 2007a, p.6)
3.4.2 Evacuation shelter accessibility
Instead of defining the cost surface as the distance between starting point and each cell in the
domain, it is possible to define it as the distance between each cell and the “Evacuation shelter
points” (which are more than one). The value of each cell is the Cost Weighted Distance (CWD)
between cell and the closest evacuation shelter point.
The process steps for the accessibility modelling will be explained as follows:
1. The Cost surface was created on the basis of land use information of the study area. Land use
data for Bali were available in a vector- format with a scale of 1 : 50 000 and had to be converted
to raster format in a first step.
36
2. Reclassification of the land use was a basic step of the analyses. Each class of the dataset got a
new value, describing their capability to modify the speed of walking person (ADPC 2007A). The
new values represent how much the average speed will be conserved on the different land use
classes. Based on a value of 100 (100% speed conservation), for example, a value of 80 means
that the speed will be reduced by 20%. Due to the fact that not every class of the dataset is a
passable one, like water bodies or mangroves, a suitable filter had to be set. For further analyses
all classes had to be retained, therefore relatively inaccessible classes got the value 1 or 5.
As presented in Chapter 3.1.3., the road network was evaluated considering the width, which was
described with number of lanes, and the traffic volume during day and night. For the accessibility
modelling, a further evaluation regarding the speed conservation had to be conducted and
combined with the land use data. The following Table 13 shows the estimated speed conservation
values of the roads in the study area.
Road_classification Width_ID 1 1 1 2 2Traffic volume_ID 1 2 3 1 2 Limitation value 50 65 80 55 70Land use_name Road_lim50 Road_lim65 Road_lim80 Road_lim55 Road_lim70 Width_ID 2 3 3 3 4Traffic volume_ID 3 1 2 3 1 Limitation value 85 60 75 90 75Land use_name Road_lim85 Road_lim60 Road_lim75 Road_lim90 Road_lim75 Width_ID 4 4 5 5 5Traffic volume_ID 2 3 1 2 3 Limitation value 80 95 85 90 100Land use_name Road_lim80 Road_lim95 Road_lim85 Road_lim90 Road_lim100 LEGEND (ID_name) Width 1 One lane small Traffic volume 1 Very much 2 One lane wide 2 Much 3 Two lanes small 3 Not much 4 Two lanes wide 5 Four lanes
Table 13 Estimated speed conservation values for the road network in the study area (Source: own composition - Data basis: own estimation)
37
The reclassified land use classes are presented in Table 14 below.
VALUE COUNT CLASS_EVAC
1 3643251 Mangrove/Lake
5 17150714 Building/River/Pond
40 5057466 Rice field
50 2008964 Dense Vegetation
65 1236 Road_lim65
70 4237 Road_lim70
75 62162 Road_lim75/Bridge_lim75
80 1376979 Open Vegetation/Road_lim80
85 301173 Road_lim85/Bridge_lim85
90 363356 Beach/Road_lim90
95 4064524 Small Vegetation/Open Field/Road_lim95
100 1086264 Bridge_lim100/Road_lim100
Table 14 Reclassified land use classes for the definition of speed conservation values (Source: own composition - Data basis: ADPC (2007a), modified)
3. The cost that must be spent to walk through each cell will be computed combining the average
speed of evacuation with the reclassified value of land use. As a statement in sec/m is more useful
for the analyses, the inverse value of this combination was calculated for the speed map. In this
research, the slowest speed of evacuees’ movement during evacuation is assigned for the
evacuation modelling. According to the Institute of Fire Safety and Disaster Preparedness Japan
(SUGIMOTO ET AL. 2003), the slowest evacuees’ speed is 0,751 m/sec that occurs on a group of
elderly walking people. The Asian Disaster Preparedness Centre (ADPC 2007A) emanate from a
value of 1,2 m/sec, which seems to be more realistic for the majority of evacuees. Therefore this
value is used for the accessibility modelling.
4. Evacuation shelter points are the basis for an evacuation time map. In this surface the value of
each cell represents the cost (in terms of time) necessary to go from there to the costless shelter
following the fastest path. For horizontal evacuation, the shelter points were placed along the
border to potential safe areas gained from inundation modelling results considering accessibility
bottlenecks (mostly big roads). Evacuation shelter points for vertical evacuation were placed
considering the evacuation shelter buildings which were identified during the field work.
The evacuation time map was calculated using the Cost allocation tool in ArcGIS on the basis of
the speed map and starting from each of the placed shelter points.
A further output of this calculation is a map with catchment areas which allows dividing the whole
space of the area into shelter sectors. The raster identifies which cells will be allocated to which
source, that means it represents the portions of the area referring to each of the shelter points.
The whole accessibility modelling process is presented in Figure 16.
38
Figure 16 Process steps for the accessibility modelling approach (Source: own illustration – Data basis: ADPC (2007a)
39
3.5 Evacuation modelling
As presented in Table 2 in the previous chapter, the ability to evacuate has to be calculated in
order to give a decision support for evacuation in case of a tsunami event. Therefore some further
information is needed as shown in the Figure 17 below.
Figure 17 Relevant information to calculate people’s ability to evacuate (Source: own illustration)
3.5.1 Shelter area accessibility for day and night population scenario
In Chapter 3.3, two methods (A and B) for a disaggregation of population data were presented.
The evacuation modelling was conducted using the results of these two methods (for day and
night population) and also with raw census data, to contrast differences between the certain
results.
40
The first modelling process step was to calculate how many people in each catchment area are
able to reach the evacuation shelter in a given time. The estimated arrival time of the used
scenario is about 30min until a tsunami wave will reach the coast of the study area. While it is
assumed that it will take about 5-10 minutes until the tsunami warning message will reach the
exposed population (GTZ- IS 2007), time limits of 15 and 25min were set by an appropriate
classification of the evacuation time map. That means that the share of the population within a
certain time zone have enough time to access the nearest evacuation shelter, outside the time
zone it is not possible.
First of all, the disaggregated population data had to be intersected with the evacuation shelter
catchment areas to calculate the particular population density and amount of people in every
potential catchment area for the identified shelters. By using the Union tool in ArcGIS, shelter
catchment area, time map and the certain population data were combined.
The population per square meter from every disaggregated dataset was multiplied with the new
areas gained from the Union- process and the sum for every catchment area was calculated. The
percentage of the population inside a defined time zone (15 or 25min) around a shelter, to the
population inside the whole catchment area, shows the particular ability to access the nearest
shelter within the stated time on the basis of the used population distribution. An example is given
in Figure 18 below.
Figure 18 Example for the shelter accessibility within 15 minutes (Source: own illustration - Data basis: own calculation; Map basis: Quickbird satellite image)
41
The conducted calculations in all combinations are presented in Table 15 below.
Population data Temporal distribution Combined Evacuation time zone
Method A Day 15min
Method A Night 15min
Method A Day 25min
Method A Night 25min
Method B Day 15min
Method B Night 15min
Method B Day 25min
Method B Night 25min
CENSUS Data Not available 15min
CENSUS Data Not available 25min
Table 15 Conducted calculations for the shelter accessibility in the study area (Source: own composition)
3.5.2 Evacuation ability for day and night population scenarios
Values for the shelter accessibility only show that a certain number of people is able to reach the
nearest shelter within a given time. However, the real ability to evacuate depends on the
accessibility and capacity of a shelter area.
The capacities of the defined evacuation shelter buildings (ESB) for vertical evacuation in the
study area could not be calculated exactly because of problems with permit procedures. Therefore
the potential evacuation space was estimated using literature results. BUDIARJO (2006) estimated
the available evacuation space of evacuation shelters by a calculation based on architectural
design space requirement. The space required for accommodating evacuees is 1m² per person. It
can be described as 0,8m² per person for stay and 0,2m² per person for circulation. Hence, the
capacity (in person unit) can be calculated as:
Available evacuation space * building area / 1
where:
Available space = available reserve space for evacuation in existing potential ESB (m²)
Building area = evacuation floor(s) area of existing potential ESB (m²)
1 = space requirement for accommodating evacuees (m²/person)
The building area was roughly estimated based on information from a satellite image and
photographs. BUDIARJO(2006) explained the assumptions and formula to calculate the capacity for
evacuation of some building functions which are potentially suitable for vertical evacuation:
42
1. Market building
Capacity for evacuation: 23% * building area / 1
Space requirement for market building design is 5.2m² per person comprises: 1m² per person for
buyer; 0.2m² per person for circulation; 2m² per person for seller and merchant display; and 2m²
per person for storage. From this space requirement, the spaces that can be occupied for
evacuation purpose are the buyer area (1/5.2 = 19.2% area) and circulation area (0.2/5.2 = 3.8%
area). Hence, the total available space for evacuation is 19.2%+3.8% = 23% of the total building
area. Regarding the crowd in this building, in evacuation situation the building can only
accommodate the people located in it.
2. Hotel
Capacity for evacuation: 26.3% * building area / 1
Space requirement for hotel design is 16m² per person comprises: 12m² per person for staying
and 1.5 m² per person for circulation; 0.5m² per person for utilities; 1.5 m² per person for hall,
lobby, and restaurant (public function); 0.3m² per person for employees’ room; and 0.2m² per
person for office function. From this space requirement, the spaces that can be occupied for
evacuation are the area (1.5/16= 9.4% area); public function area (1.5/16 = 9.4% area); and
assumed 10% non-occupied rooms (0.1*12/16 = 7.5% area). Hence, the total available space for
evacuation is 9.4%+9.4%+7.5% = 26.3% of the total building area.
For the study area some other building functions were suitable for vertical evacuation. The
potential evacuation space was estimated based on information from interviews and the inspection
of similar buildings.
3. Storehouse
Capacity for evacuation: 25% * building area / 1
Space requirement for storehouse is estimated based on a building observation during the field
work.
4. Billiard Centre
Capacity for evacuation: 50% * building area / 1
Space requirement for storehouse is estimated based on a building observation during the field
work.
The evacuation space within safe areas for horizontal evacuation was calculated on the basis of
suitable (passable) land use classes. The following classes were selected:
(1) Small Vegetation
(2) Open Vegetation
(3) Open field
43
The calculation for the evacuation ability for day and night population scenarios is based on the
same principle as for the shelter area accessibility in Chapter 3.5.1.
The difference is that the population inside a defined time zone (15 or 25min) around a shelter has
to be compared with the shelter capacity. If the shelter capacity is smaller than the number of
people which are potentially able to evacuate, the value for the real Evacuation ability is smaller
than the value of the Accessibility. If the shelter capacity is bigger than the population within the
evacuation time buffer, the value for the Evacuation ability equates to the value of the
Accessibility. An example is given in Figure 19.
Figure 19 Example for the evacuation ability within 15 minutes (Source: own illustration - Data basis: own calculation; Map basis: Quickbird satellite image)
45
4 Result
4.1 Data collection
Identification of the road network
As described in Chapter 3.1.3, all roads in the study area were evaluated considering their width
and the traffic volume. The results are presented in Figure 20 and 21. It becomes evident that
most of the roads are single- lane and only a few main road axis cross the area. Regarding the
traffic volume it is obvious that there is a strong dependence on the road width. The bigger the
road, the more traffic can be noticed. Exceptions are touristic and commercial areas, where
smaller roads are also highly frequented (cp. Figure 22).
An important observation, especially for pedestrian evacuation, is the bad road infrastructure near
the coast. Only a few roads in the North and in the centre of the study area lead directly to the
beach, whereas in the South only some small ones connect the beach area with the main road
infrastructure. Main traffic junctions are located in the eastern centre and in the South of the study
area.
Figure 20 (left) /21 (right) Traffic volume (left) and road width (right) in the study area (Source: own illustration – Data basis: own data collection; Map basis: Quickbird satellite image)
46
Figure 22 Functional urban sectors in the study area (Source: own illustration - Data basis: own data collection; Map basis: Quickbird satellite image)
47
Identification of functional urban sectors
The sectors were identified according to the process presented in Chapter 3.1.3. Figure 22 shows
the gained zonation of the study area. The functional structure defines a very clear separation
between the local and the touristic population. Residential areas of the tourism sector (Hotels,
Bungalows, etc.) are mostly located near the coast, touristic commercial areas along the main
roads in the northern part of the study area. The areas of the local population are located further
inland, business areas mostly along the main roads in the southern part. Nightlife districts are
concentrated in the centre of the study area and along the main road in the tourism sector.
Common functional sectors of the touristic and the local population are visible in the centre of the
area, mostly commercial districts, like shopping centres, supermarkets and food stores.
Identification of evacuation shelters and critical facilities
Evacuation shelters were identified considering building functions potentially suitable as evacuation
shelter building, general characteristics of suitable shelters from literatures and required additional
criteria which were find out by expert interviews and technical literature (BUDIARJO 2006, EISNER
AND NTHMP 2001) (cp. Chapter 3.1.3 ). The buildings are presented in Table 16 below.
No Name,
location,
Function
Shore distance
Evacuation type
Shelter accessibility
Number of Storeys, Evacuation space capa-city, Space accessibility
Map Image Image
1 Hotel Arena,
Kelurahan Legian,
Hotel
350m
Vertical evacuation
Good 2 storeys
Big
Good
2 Paradiso Billard,
Kelurahan Kuta,
Billard Centre
400m
Vertical Evacuation
Good 2 storeys
Big
Good
3 Hotel Kuta Royal,
Kelurahan Kuta,
Shopping Centre
900m
Vertical Evacuation
Good 1 storey
Big
Good
4 Supernova,
Kelurahan Kuta,
Shopping Center
1100m
Vertical evacuation
Good 1 storey,
Big
Good
Not available
5 Pepito storehouse,
Kelurahan Kuta,
Storehouse
1200m
Vertical Evacuation
Good 1 storey
Big
Good
Not available
6 Hotel Jatra,
Kelurahan Kuta,
Hotel
500m
Vertical Evacuation
Good 2 storeys
Big
Good
7 Pasar Kuta,
Kelurahan Kuta,
Market building
800m
Vertical Evacuation
Good 1 storey
Big
Good
Not available
8
Hotel Baru,
Kelurahan Seminyak
Hotel
300m
Vertical Evacuation
Good 1 storey
Big
Good
Not available
Table 16 Potential ESB’s in the study area, identified during the field work (Source: own composition)
50
Figure 23 Critical facilities in the study area (Source: own illustration - Data basis: own data collection; Map basis: Quickbird satellite image)
51
Critical facilities were also located during the field work. Number and position of hospitals and
schools could be found out from the Department of Health and Education in the district of Badung.
Further relevant information, like number of students or patients, building size, etc., were not
available and had to be asked in the certain facilities. However, a questioning of hospitals in the
study area was not successful, since the rate of return of the questionnaires was very bad until the
end of the field work. As shown in Figure 23 most of the hospitals are located in the eastern part
of the study area and more than 1 kilometre away from the coast. A good accessibility to main
evacuation routes is ensured since nearly all facilities are located along main roads. All buildings
are at least two- storeyed and therefore the basic potential to evacuate people to higher floors is
given.
Further information relevant for evacuation planning could not be gained during the field work.
The schools are widespread in the study area. Most of the facilities are located within a distance of
less than 1 km from the coast and are only accessible by smaller roads. The questioning of the
schools in the study area was more successful and the following additional information could be
gained during the field work:
No. Number of storeys Number of students Schooltime 1 1 261 7.30 - 12.00h 2 1 536 7.30 - 12.00h 3 2 2124 7.30 – 12.00h
13.00 – 18.00h 4 2 1183 7.30 – 12.00h
13.00 – 18.00h 5 1 346 7.30 – 12.00h
13.00 – 18.00h 6 1 85 7.30 – 12.00h
Table 17 Questioning results from the school sector (Source: own composition)
4.2 Hazard impact and exposed area
As described in Chapter 3.2.1, both historical records of tsunami impact observations and tsunami
inundation modeling results were the basis to conduct a hazard assessment for the study area. In
the following the results of both methods are presented.
Using historical records of tsunami impact observation
A first tsunami impact zone classification can be mapped based on distance zones and their
probability to be affected by a tsunami. The result map (see Figure 24, left) shows different
frequency rates of an event. The zonation is mapped in 100m steps until a distance of 1 km and
further in 1 km steps until a distance of 10000m from the coast. Although, the inundation zones of
more than 60 % of all tsunami events were less than 100 m inland (own calculation), it is obvious
52
that the probability to be affected by a tsunami wave is very high for the study area. The next
important parameter is terrain elevation above sea level, because the influence of altitude on
tsunami inundation is obviously an important factor (POST ET AL. 2007). The altitude zonation also
takes into account river valleys, where the tsunami can run further inland. These valleys are highly
exposed areas. The classification of elevation classes strongly depend on the resolution of the
respective elevation data used. Because of the high resolution of the Digital Surface Model (DSM),
the zones are mapped in 1m steps until an elevation of 15m, in 5m steps until 50m and in 10m
steps until an elevation of 80m (see Figure 24, middle). The zone with a very high probability of
being affected by a tsunami reaches from 0 to 10m above sea level. The study area is a very flat
area and therefore, with a high probability, completely affected by a possible tsunami event.
Beside historical information on tsunami events, elevation and distance relationships, “protective
measures” have to be considered as well (POST ET AL. 2007). Hence, a distance- elevation
relationship in combination with surface roughness (land cover data) can be applied to develop
tsunami impact zones. The results of these zonations are presented in Figure 24 (right). It
becomes evident, that protective measures in the study area mitigate the tsunami intensity. Due
to the urban structure it is assumed that mainly buildings cause the mitigation effect.
Figure 24 Tsunami hazard zonation of South Bali, based on historical records of tsunami impact observation. From left: (a) probability of affectness regarding the horizontal distance, (b) probability of affectness regarding elevation, (c) probability of exposure as a combination of distance, elevation and surface roughness (Source: own illustration - Data basis: NGDC; Map basis: Quickbird satellite image)
Using inundation modelling results
As described in Chapter 3.2.1, a multi- scenario approach analysis the likelihood of tsunamis of
various sizes that can then be simplified into tsunami hazard zones. Altogether 372 Scenarios were
considered in the analyses for the South of Bali in Figure 25. The scenarios are distributed equally
within the tsunamigenic sources in the Sunda Trench region (impacting coastline of Sumatra, Java,
53
Bali), where seismic activities may introduce ocean bottom deformations and consequently
tsunamis. The scenarios reflect several possible tsunamigenic earthquakes with moment
magnitudes of 8.0, 8.5 and 9.0. The hazard impact zones reflect the percentage of all scenarios
which inundate a particular land area. Hence for the red zone, between 70% and 95% of all
scenarios impact the coastland. The yellow zone (moderate impact) denotes up to 10% to 70%,
the green zone 3% to 10% impact of all scenarios. Looking at the results in Figure 25, it becomes
evident that the study area is completely affected by 10% to 70% of all available scenarios.
Figure 25 (left) / 26 (right) Hazard impact zones in the South of Bali, based on a multi- scenario approach (left) and on one inundation scenario (right) (Source: own illustration – Data basis: AWI; Map basis: Quickbird satellite image
However, the identification of potential safe areas within the study area is essential for further
analysis regarding accessibility and evacuation modelling. Therefore, for the study area only one
scenario is considered as a basis for the modelling process in this research as presented in Figure 26. The scenario shows that flow depth values between 3.0 and 10.0 m only occur as run- up at
the coastline while the smaller depths until 3.0m show the impact at the coastland. Most of the
study area is affected except some smaller areas in the North and in the South.
54
4.2.1 Potential safe areas
As presented in Chapter 3.2.2, areas which are not affected by a tsunami wave and hence outside
the inundation zone, can be regarded as potential safe area.
Figure 27 (left) / 28 (right) Safe areas in the study area based on historical tsunami data (left) and on one inundation scenario (right) (Source: own illustration – Data basis: NGDC; Map basis: Quickbird satellite image)
The exposed areas gained from historical data are based on the probability of affectedness relating
to the historical records of tsunami impact information. To derive safe areas, a very low probability
value has to be defined. Therefore a minimum value of 10% is defined to identify areas as safe
respectively outside potential inundation zones (EXPERT INTERVIEW DR. KAI ZOßEDER, 23.04.2008).
Considering the further requirements for safe areas to ensure a good accessibility and a sufficient
capacity for a huge amount of evacuees (Area > 10000m² and Slope < 20°) as presented in
Chapter 3.2.2, safe areas for the study area were derived accordingly (see Figure 27). It becomes
evident, that except some small parts in the North, the whole study area is located within the
calculated hazard impact zone. Therefore this result is not suitable for further analysis where safe
areas within the study area are essential.
Regarding the multi- scenario approach it is already mentioned that up to 70% of all available
inundation scenarios completely affect the study area and no safe areas can be derived. Therefore
it is necessary to use the artificial scenario for further accessibility and evacuation modelling. As
55
shown in Figure 28, the two areas in the South and in the North of the study area are outside
inundation zone and meet the additional requirements for safe areas (cp. Chapter 3.2.2).
4.3 Population modelling
As described in Chapter 3.3, it is quite difficult to disaggregate population census data because
adequate knowledge of the location of the dwellers is necessary, as there can be a high variation
in area, number and characteristics of the population.
The conducted population disaggregation in the study area using two different methods is based
on different disaggregation units and spatial resolutions.
Method A: With the combination of land use and census data, population data can be
disaggregated to assigned land use types as the smallest unit (KHOMARUDIN ET AL. 2008). Land use
data were available in a scale of 1 : 50000.
Method B: Population data can be disaggregated to one functional urban sector with the
combination of remote sensing data and field work results.
The Figures 29/30 and 31/32 below show the day and night population in the study area
calculated by Method A and B. As it can be seen in the Figures 29/30, the population data are
represented by larger spatial disaggregation units as in Figure 31/32, which is a result of the lower
spatial resolution of the input data.
Weighting factors considering the degree of people activity in every land use class as well as the
number of people in the input census data, represent the disaggregated population distribution.
The land use class which covers the study area is mostly settlement area and it is assumed that
this class represents a very high activity (KHOMARUDIN ET AL. 2008). Two main results can be
derived for Method A:
→ Both for the day and the night population, the study area is divided into four zones with a
different number of people. This is only a consequence of the used input data (Land use classes
and census districts) and cannot be related to the real population distribution in the urban
agglomeration of the study area.
→ Regarding the differences between the day and night population it can be seen that during the
night there are more people in every zone than during the day. Due to this common increase, it
becomes evident that no population movements between the different zones are considered, but
only an increase of people from outside the study area.
56
Figure 29 (left) / 30 (right) Distribution of the day and night population in the study area, calculated with Method A (Source: own illustration - Data basis: Khomarudin et al. 2008; Map basis: Quickbird satellite image)
Figure 31 (left) / 32 (right) Distribution of the day and night population in the study area, calculated with Method B (Source: own illustration - Data basis: own data collection; Map basis: Quickbird satellite image)
57
Method B in Figure31/32 is based on functional sectors gained by field work. Sectors of interest
are city districts which show a similar structure of building functions and from which people
activities during different times of day and night can be derived. In this context, the tourist
population plays an important role because of the strong dominance of touristic infrastructure in
the study area. Regarding the results, population hotspots become evident on a first view. A clear
difference between the areas near the coast and the hinterland are visible as marked with a line
for the day population in Figure 31. To a large extent, the hinterland is more populated both
regarding the day and night population. The most densely populated sectors are located in the
North- West and in the centre of the study area.
In comparison with Method A, the results provide a more differentiated insight into the population
distribution of the study area. The spatial disaggregation units are much smaller and independent
from based input data, as the census districts in Method A. In Figure 33 and 34, a sample area is
separated to compare the number of people calculated with Method A and B and to get an idea
how far both methods are suitable for further analysis in this research. The population values
differ widely from each other and the limited significance of the disaggregation approach with
Method A becomes evident. Therefore this Method is unaccounted for the next modelling steps as
more detailed population data are required for evacuation planning.
Figure 33 (left) / 34 (right) Comparison of the population number in two sample areas based on the calculations with Method A and B (Source: own illustration - Data basis: own calculation; Map basis: Quickbird satellite image)
58
In the following, the results from Method B will be analysed regarding population movements
between the functional urban sectors (cp. Figure 35).
Basically, the population distributions show considerable differences between day and night.
During the day, commuter movements to the workplaces in the study area are considered which
leads to increasing values in the business areas for the local and tourist population. Generally the
population activity is more concentrated on commercial areas and beach areas both for local
people and tourists, while the activities in the residential areas are weighted less. The highest
population concentration can be stated in areas where local and touristic population have common
activities, as can be seen in the mixed commercial areas.
Figure 35 (left) / 36 (middle) / 37 (right) Functional urban sectors in the study area (left); distribution of day (middle) and night (right) population in the study area, calculated with Method B (Source: own illustration - Data basis: own data collection; Map basis: Quickbird satellite image)
The night population is more concentrated on residential areas both for the tourists and the local
population, as well as in the nightlife districts, where the highest activity is assumed. The lowest
concentration can be stated again in the vegetation areas and also in the commercial areas for
local people, because there is no activity assumed outside the working time.
In Table 18 the calculated day and night population of each functional sector is presented. The
values confirm the visual interpretation of the maps. During the day the beach and the commercial
areas both for the tourism and the local sector are highly frequented. At night, most of the people
stay in the residential or in the nightlife areas.
59
Functional type Number of people - Day Number of people - Night
Beach 922 82 Business mixed/Nightlife 1303 2018
Business mixed/Shopping 400 71 Local- business area 3749 209
Local- residential area 11354 18225 Local- mixed area 809 1039
Tourism- residential area 1179 1967 Tourism- business/Nightlife 886 1343
Tourism- business/Shopping 2135 536 Tourism- mixed area 1117 758 Tourism/Local- mixed 620 300
Table 18 Differences between the day and night population calculated with Method B (Source: own composition – Data basis: own population disaggregation concept)
4.4 Accessibility modelling
The estimated evacuation speed depending on the characteristics of all passable paths is the basis
of the shelter accessibility modelling. As described in Chapter 3.4.2, appropriate values were
assigned to the land use classes considering the particular speed impedance of each class. In
Table 19, the time required to walk through one cell (1m) of each land use class is presented.
Land use class sec/m Land use class sec/m Beach 0,93 Road_lim60 1,39
Building 16,67 Road_lim65 1,28 Dense Vegetation 1,67 Road_lim70 1,19
Lake 83,33 Road_lim75 1,11 Mangrove 83,33 Road_lim80 1,04 Open Field 0,88 Road_lim85 0,98
Open Vegetation 1,04 Road_lim90 0,93 Pond 16,67 Road_lim100 0,83
Rice Field 2,08 Bridge_lim75 1,11 River 16,67 Bridge_lim85 0,98
Small Vegetation 0,88 Bridge_lim100 0,83 Road_lim55 1,52
Table 19 Time required walking through one cell (1m) of each land use class (Source: own composition)
60
Most of the values range between 0,8 and 1,5 sec/m, which is a realistic variation from the based
average speed of 1,2 sec/m for pedestrian evacuation (ADPC 2007A). The very high values (e.g.
for Lake and Mangrove) are the result of the estimated high impedance values. Actually, these
land use classes are not passable but had to be considered in the calculation to provide an are-
wide cost surface for the study area. In Figure 38 and 39, the land use classes in the study area
and the inverse evacuation speed per cell are contrasted. The dark areas show land use classes
with a high speed impedance, mostly buildings. It is assumed, that it is possible to go through a
building, for example using the front and the back door, but it takes a lot of time. The road
network has obviously low impedance values and seems to be the best evacuation path because of
its dense network. But it becomes apparent that also other land use classes show very low speed
values. They often don’t have connections to each other but they function as important interfaces
between the road networks as the preferred evacuation routes. For example, the marking A in
Figure 39 shows open field areas which are important connections between the safe area in the
South and the surrounding road network. Marking B also shows an important connection between
the beach area and the safe area in the North, consisting of small and open vegetation. All well-
passable land use classes are an important improvement of the common evacuation road network.
Figure 38 (left) / 39 (right) Left: Land use classes in the study area (Source: own illustration – Data basis: LAPAN; Map basis: Quickbird satellite image; Right: Evacuation speed per cell in second per meter (Source: own illustration – Data basis: own calculation; Map basis: LAPAN land use classes
A
B
61
Based on the evacuation speed map, the shelter accessibility is calculated. Figure 40 shows the
evacuation time (in min) from every point (cell) in the study area to the nearest (in terms of cost)
evacuation shelter. The shelter catchment areas mark the accessibility borders to the several
shelters.
The road network is in evidence and mark mostly short access times. Other areas, although they
are near an evacuation shelter, have higher time values because of a relatively high impedance of
the underlying land use class. The map shows that all people are able to evacuate within 40min,
based on the estimated cost values for each cell. Looking at the markings A and B in Figure 40, it
becomes evident that there are some critical areas along the coast with very high evacuation
times. For evacuation planning it is of great importance to identify such evacuation hotspots with
bad evacuation shelter accessibility options, especially in areas where population density is high.
The accessibility modelling approach in this research provides a tool to define the fastest
evacuation route from a source point to the nearest shelter in order to clearly identify the
evacuation possibilities of critical areas.
Figure 40 (left) / 41 (right) Left: Evacuation shelter accessibility based on a CWD approach (Source: own illustration – Data basis: own calculation; Map basis: inverse evacuation speed map); Right: Example for the definition of “fastest evacuation routes” (Source: own illustration – Data basis: CWD surface; Map basis: population distribution, calculated with Method B)
62
In Figure 41 the fastest evacuation routes from some source points to the nearest evacuation
shelter are presented exemplarily. The example routes show that the fastest path is not always the
shortest path (which is the direct line between the source point and evacuation shelter).
The accessibility modelling, as presented in Figure 40, is based on defined evacuation shelter
points: (1) Shelter access points for horizontal evacuation and (2) evacuation shelter buildings for
vertical evacuation. While the horizontal access points are only based on one potential inundation
scenario and the vertical shelters are defined in one’s sole discretion, it have to be made clear that
the modelling is only a pre-operating approach. The buildings were defined based on known
criteria (cp. Chapter 3.1.3) gained from literature (BUDIARJO 2006, EISNER AND NTHMP 2001) and
new expertises during the field work process. The inundation scenario, as the basis for the assign-
ned safe areas for horizontal evacuation, was chosen because of the good representation of
relevant characteristics of a possible tsunami event which could occur the study area. Generally,
the great importance of vertical evacuation within the study area is beyond all questions. In a
coastal urban area where population and building densities are high, vertical evacuation to a
reachable evacuation shelter can be considered as the most appropriate action. In these
circumstances, evacuating people to the second or third floor of robust buildings which withstand
the earthquake and tsunami waves is the most effective way to save lives (BUDIARJO 2006). In
Chapter 3.1.3, general and specific requirements for evacuation shelter buildings are presented
with the advice to a necessary professional technical expertise for an official assignment of
buildings as evacuation shelters. Due to a lack of time and technical knowledge, such an
inspection was not feasible in this study. Regarding all general, specific and technical requirements
for an evacuation shelter building, a criteria check list presented in Figure 42 was developed and
can be used for further studies.
63
64
Figure 42 Criteria check list for tsunami evacuation shelter buildings (ESB) (Source: own illustration – Data basis: BUDIARJO 2006, EISNER AND NTHMP 2001)
65
4.5 Evacuation modelling
4.5.1 Shelter accessibility for day and night population scenario
As described in Chapter 3.5.1, the first step of the evacuation modelling process is to calculate
how many people in each catchment area are able to reach the evacuation shelter in a given time
considering three different population distributions (Method A and B, census population) and two
different time limits (15 and 25min). In Figure 43, 44 and 45, the shelter accessibility within a time
limit of 15min is presented for all population distributions (day situation).
Figure 43 (UL) / 44 (UR) / 45 (LL) Evacuation shelter accessibility for the day population (calculated with Method A, B and the census population) within 15 min – (Source: own illustration - Data basis: CWD surface; Map basis: catchment areas of the evacuation shelters, Figure 29, 31 and census population map)
66
The value 0 shows the worst, the value 1 the best evacuation shelter accessibility. By a first visual
interpretation of the results, it becomes evident that the accessibility values for the calculation with
Method B considerably differ from Method A and the census population. The latters show a better
accessibility for many catchment areas, whereby the results among each other are very similar.
For all different population distributions the accessibility values generally are very high. Figure 40
can explain this circumstance, because the map shows that nearly all shelters are reachable within
15min. The results of the calculations with a time limit of 25 show nearly the same differences and
are presented in Appendix 4 and 5.
Table 20 itemizes the calculated results regarding the shelter accessibility of each catchment area,
to offer a more detailed interpretation. Comparing the different accessibility values, the statement
of the visual interpretation can be confirmed but the differences between the three Methods are
marginal. The percentage of the people who are able to evacuate within 15 minutes to the total
population in each catchment area is similar for each method, although the population values are
very different. Evacuation shelter number 2 shows the worst accessibility values in all calculations
while the access points to safe areas for horizontal evacuation are reachable for most of the
people.
Table 20 Calculation results for the shelter accessibility of each catchment area (Source: own composition – Data basis: own calculation)
67
The Figures 46/47 and 48/49 show the shelter accessibility for the day and night population of
Method A and B within a time limit of 15min. According to the population distribution in Chapter 3.3, only the calculation with Method B show obvious differences for the day and night
accessibility, while for Method A the values are very similar. In general, both calculations show
that a majority of the population in each catchment area is able to access the nearest evacuation
shelter both during day and night.
Figure 46 (left) /47 (right) Evacuation shelter accessibility within 15 min for day/night population, calc. with Method B (Source: own illustration - Data basis: CWD surface; Map basis: catchment areas of evacuation shelters, Fig. 31 and 32)
Figure 48 (left) /49 (right) Evacuation shelter accessibility within 15 min for day/night population, calc. with Method A (Source: own illustration - Data basis: CWD surface; Map basis: catchment areas of evacuation shelters, Fig. 29 and 30)
68
Table 21 and 22 confirm these visual impressions. Looking at Method B, the evacuation shelters 1,
2, 5 and 21 show the biggest differences between day and night while for the shelters 8 , 9, 10,
11, 12 and 15 no differences are statable. Regarding the accessibility values of Method A, only
shelter 1, 19 and 20 shows very small difference values, for all other calculations no differences
are identifiable.
The results will be further discussed in Chapter 5.4.
Table 21 (left) / 22 (right) Difference between day and night accessibility values, calculated with Method A (left) and B (right) (Source: own composition – Data basis: own calculation)
4.5.2 Evacuation ability for day and night population scenario
The evacuation ability is a result of the accessibility considering the capacities of the certain
evacuation shelters. Hence, the number of people who is able to evacuate is limited to the certain
capacity of the evacuation shelters.
For the calculation, the blue safe area access points are treated as one shelter point for each safe
area. That means that the six access points for the upper safe area and the seven access points
for the lower safe area are considered in each case as only one shelter point in the calculation for
the evacuation ability map ( as signed with number 12 and 21 in the Figures 50/51 and 52/53).
This modification of the input parameter was conducted because for the calculation of the
evacuation ability, the shelter capacity is of importance. The mentioned access points are related
to only two safe areas and therefore were combined accordingly. The results show again that the
day and night results calculated with Method B differ more from each other than the results of
Method A. It is also obvious that the evacuation ability of most of the catchment areas differ
widely from the related accessibility values which do not consider shelter capacities.
69
Figure 50 (left) /51 (right) Evacuation ability within 15 min for the day and night population, calculated with Method B (Source: own illustration - Data basis: CWD surface; Map basis: catchment areas of the evacuation shelters)
Figure 52 (left) /53 (right) Evacuation ability within 15 min for the day and night population, calculated with Method B (Source: own illustration - Data basis: CWD surface; Map basis: catchment areas of the evacuation shelters)
Generally, the Figures 50/51 and 52/53 show very different values. Table 23 and 24 itemize the
calculated results regarding the evacuation ability of each catchment area. Especially shelter 5 and
8 show considerable different evacuation ability values in Method A and B. Regarding the day
situation for shelter 8, Method A calculates a value of 0,36 and Method B a value of 1,0.
70
Also the calculations for shelter 5 show different results, while the values range between 0,67
(Method B) and 0,92 (Method A) for the day population. This is a consequence of the different
population values. While the particular evacuation shelter capacities are the same for both
calculations, the population values differ widely from each other. In Method A, for example, 414
people are able to reach Shelter 8 within 15 minutes, but due to the shelter capacity of 151 people
only 36% of the people can be evacuated. In Method B, only 141 people are estimated within a
time buffer of 15 minutes around the shelter and therefore all people are able to evacuate. The
good values for the catchment areas of the safe area access points, both in Method A and B, are a
result of the sufficient capacities of these areas.
The results will be further discussed in Chapter 5.5.
Table 23 Difference between the evacuation ability of the day and night population, calculated with Method A (Source: own composition – Data basis: own calculation)
Table 24 Difference between the evacuation ability of the day and night population, calculated with Method B (Source: own composition – Data basis: own calculation)
71
5 Discussion
5.1 Data collection
The availability of suitable data is the basis of evacuation modelling. The data research in Bali was
a difficult process due to different factors: (1) To get statistical data and to interview experts, the
support of responsible administrations was essential. However, the willingness to cooperate was
often not high and therefore very time- consuming. (2) Another point was the unclear
responsibilities between the local administration among each other and between the local and
regional administration. Due to these problems, the data research had to be restricted in some
cases. In the following, all data which were essential for this study are presented in Table 25 regarding their availability. In addition, further data which could be used to broaden and to
improve this approach are attached. These attached data were not considered in this thesis
because of the limited frame of the research approach or because of a lack of time as a result of
the above mentioned problems.
No. Data requirement Availability and data source
Additional useful data
1. Tsunami Hazard Parameter:
a) Historical data (horizontal distance of inland flooding and flow depth)
b) Numerical inundation models
a) available, derived from NDGC
b) available, provided by AWI
- Complete historical data base for Indonesia (NDGC data are incomplete)
- High- resolution DEM to derive detailed slope values
2. Street network
a) Street width
b) Traffic volume
Digital street data provided by (BAKOSURTANAL)
a) available, gained by field work
b) available, gained by field work
- Street surface (pavement, condition)
3. Land use data Available, provided by LAPAN - Digital building mask
4. Evacuation shelter locations
a) for vertical evacuation
b) for horizontal evacuation
a) available, gained by field work
b) available, derived from inundation modelling
a) detailed building data of potential ESB (Structure, design, capacity, accessibility)
5. Location of critical facilities Partly available, gained by field work (schools, hospitals)
- Detailed data of critical facilities (Number of people, storeys, accessibility, capacity)
6. Detailed population data Available, gained by field work
- Building- based population data
- Socio- economic data
- Detailed tourist data
- Commuters data
Table 25 Data requirement and additional useful data for the thesis (Source: own composition)
72
Tsunami hazard parameter
Tsunami hazard maps show areas that could be inundated by a tsunami. The hazard maps and
information mainly serve to educate and prepare the public and assist emergency responders to
plan evacuations (EMERGENCY MANAGEMENT AUSTRALIA 2005). For the study area both historical data
and numerical inundation models were available. While results obtained by numerical inundation
modelling possess a certain degree of uncertainty (due to incomplete knowledge on model
parameter, input data assumptions, etc.), historical event data can be used for validation. As
presented in Chapter 3.2.1, a hazard assessment was conducted, based on data from the NDGC.
However, this historical data base is incomplete. By a comparison with other sources for historical
data, it became obvious that some tsunami events are missing. A complete data base would
provide a more suitable validation possibility.
A further useful parameter for a hazard assessment would be the consideration of slope values. On
the one hand depressions and elevations which strongly influence the propagation of a tsunami
wave could be highlighted. On the other side, extreme slope values can influence the evacuation
speed. A DSM (Digital Surface Model) was available for Bali, but because of the completely flat
study area, slope values were not considered.
Street network
The street network, as the most important evacuation route in case of a tsunami, is a main com-
ponent in evacuation modelling (CHURCH AND SEXTON 2002). A digital dataset was provided from
BAKOSURTANAL and, as presented in Chapter 3.1.3, was valuated during the field work consi-
dering the width, which was described with number of lanes, and the traffic volume during day
and night. Since the quality of an evacuation route is a main factor during mass evacuation, a lot
more street characteristics could be taken into consideration for evacuation modelling. The street
surface plays an important role, while, for example, a bad pavement as well as a bad condition,
can influence the evacuation speed. The better and extensive the input data, the better the
evacuation modelling results.
Land use data
The available land use data were initially the basis of the calculation for the Cost Weighted Dis-
tance (CWD) as presented in Chapter 3.4.1. Therefore the available data were sufficient, because
the cost surface was only differentiated regarding different land use types. More detailed data
were necessary to disaggregate census population data. In Chapter 3.3 a method for population
disaggregation combining land use and census data is presented. The gained results show only a
low level of detail at the considered scale (see Chapter 4.3). A digital building mask would provide
more possibilities to downscale population data. An accordant dataset was provided by BAPPEDA
in the course of the field work, but the data were incorrect regarding topology and geometry. An
adoption of available data or the generation of new one’s, was not possible due to a lack of time.
73
Evacuation shelter locations
As presented in Chapter 3.1.3, potential evacuation shelter buildings (ESB) were identified during
the field work and evaluated regarding the public function, number of storeys and potential
capacity. Safe areas for horizontal evacuation were assigned according to the inundated areas
based on the used tsunami inundation scenario. A central part of the shelter identification was the
effort to integrate the hotel sector in this process. Due to a bad response rate of the
questionnaires, many relevant data for vertical evacuation had to be estimated. The role of the
hotel sector in evacuation planning will be discussed more detailed in Chapter 5.4. However, for a
professional identification of ESB, more detailed building information (building structure, design,
capacity, accessibility, etc.) are required as shown in the criteria check list for ESB presented in
Chapter 4.4. For the study approach used in this work, the available data gained by field work
were sufficient.
Location of critical facilities
As presented in Chapter 3.1.3 , schools and hospitals in the study area were identified during the
field work to highlight facilities of vulnerable groups, which need special attention in case of an
evacuation. Due to a lack of necessary secondary data for a detailed evaluation of the facilities, a
survey was conducted based on standardized questionnaires. Thereby, a decisive point was the
difference between public and private institutions, both in the school and in the health sector. The
public institutions were willing to provide information after presenting a recommendation from the
superior authority. From the private sector only a few information could be received. Therefore, a
detailed examination of the critical facilities was not possible.
Although, the gained results during the field work allow some basic statements regarding
evacuation planning. Four schools are one- storied and accommodate up to 500 people during the
school hours, mainly elementary school students which need special guidance in case of an
evacuation. Only two schools are multi- storied (cp. Table 17) and can potentially use the upper
floor for vertical evacuation. However, more detailed information is required to provide special
evacuation guidance for such facilities. Of great importance is to know whether the buildings meet
the general requirements for ESB. Evacuation planers have to know if the students are able to
evacuate themselves to an upper floor or if they have to move from this place to reach an
evacuation shelter. Critical facilities are mostly population hotspots (e.g. schools during school
hours or big hospitals) and therefore have to be considered in ESB location planning to ensure an
optimized shelter distribution. In case of an evacuation, it plays an important role if about 2000
students (cp. Table 17) have to move or not. Furthermore it is essential to know if, especially
elementary schools students are sensitized for the tsunami danger and if special evacuation plans
are available. It is essential to provide special evacuation guidance for such facilities. Otherwise, a
tsunami warning will cause panic and respectively uncontrolled evacuation behaviour.
74
Regarding the hospitals in the study area, the situation is more complicated. Comparing the
location of the facilities with the potential inundation zone based on the used inundation scenario
(cp. Figure 23 and 26), it is obvious that some of them are located within the identified tsunami
hazard zone. An important required information is once more if the facilities can accommodate the
people inside in an upper floor and if a special evacuation plan is available. Disabled people mostly
are not able to evacuate themselves and therefore the response time in case of a tsunami warning
is much lower than, for example, for schools. Detailed evacuation guidelines are required to ensure
the evacuation of the patients in a very short time.
While the consideration of critical facilities is such an important factor in evacuation planning, it
must be point out that in reality a lot more facilities should be implemented in an evacuation
planning process. Following a definition issued by the German Federal Ministry of the Interior
(BUNDESMINISTERIUM DES INNEREN 2005), critical infrastructures are organizations and facilities
having high relevance for the national community whose disturbance or failure would have lasting
effects on supply, considerable disruptions of public safety, or other significant adverse impacts
(e.g. high rate of loss of life).
All critical facilities are characterized by their importance for an efficient society and by their high
exposure to a tsunami. Among other things, they are further sub- divided to so called “essential
facilities” (STRUNZ ET AL. 2008B) featuring particularly endangered people (young, old, ill, disabled)
including schools, hospitals, kindergartens, residential homes for elderly or disabled people.
While the evacuation of people in risk areas within a certain time is the main concern in evacuation
planning, these facilities are considered in this pre- operating study. However, other types of
facilities are also relevant. “Supply facilities” like airport, harbor, police stations, fire brigade and
military, are important in providing general supply functions. “High loss facilities” like industry,
power plants and oil tanks, are facilities with high danger in causing negative effects for people
and environment (secondary hazards like fire, oil spill). All these facilities have different demands
for evacuation planning. Providing information about this infrastructure is important for official
evacuation planning in order to be able to define local “hotspots” for early warning and special
evacuation requirements.
Population data
As presented in Chapter 3.1.3, the sub-division of the study area into functional urban sectors was
necessary to get a more detailed impression of the population movement within one day.
Population data were gained by disaggregating an estimated total population consisting of census
data, tourist population and commuters. Functional urban sectors were defined and population
movements between these sectors were assigned considering day and night population. A main
concern of this approach was to show the different spatial behavior patterns of the tourist
population during one day. However, due to insufficient data, the tourist population had to be
estimated. A main problem was the administrative border of the study area (cp. Chapter 2.1). Due
75
to the intersection of three village borders, the allocation of statistical data to the study area was
difficult. Detailed attendances, accommodation occupancy rates and the average length of stay are
required data to improve the population data.
From the local administration responsible for tourism, no detailed data were available. As already
mentioned in Chapter 3.1.3, a hotel questionnaire was developed in cooperation with the
“Indonesia Hotel and Restaurant Association Badung Regency of Bali” to gain information about
criteria for evacuation shelters. In this framework it was also asked for tourist data, but the
response rate of this questionnaire was under 1%. Due to the dense hotel structure (including
bungalow resorts) of the study area, building based data could provide more detailed information
regarding population movements between the different functional sectors.
Also data about the number of commuters in the study area during one day would be helpful to
get a better picture of the population distribution or rather density hotspots at a particular time. It
is difficult to measure these population streams because a lot of people are working in the informal
sector and data about the residence of people working officially in the study area are not available.
The used estimation (for the commuters) is only a result of some interviews and field observation.
Altogether, the estimated total population number is disaggregated to functional sectors as the
smallest unit. Detailed building- based data would provide much better information (TAUBENBÖCK ET
AL. 2007). Area- wide data about building functions could highly improve the functional zoning.
Socio- economic data could be used to highlight vulnerable groups which need special attention
during an evacuation (see Chapter 3.1.3).
5.2 Hazard impact & exposed area
Aim of the hazard assessment is to achieve a better understanding of tsunami hazard and the
possible impacts on society in order to enable local and national decision makers and other
stakeholders to get better prepared for future tsunami events (GONZÀLEZ ET AL. 2005)
As described in the Chapters 3.2.1 and 4.2, three basic approaches are generally available to
conduct such an assessment: (1) using observations of historical tsunami events which occurred in
the respective area, (2) by using numerical inundation modelling results (one or many scenarios)
and (3) by using a combination of historical data and inundation modelling results.
For the study area the first method was used at the beginning of the research because of absent
inundation modelling results in order to create a basis for the evacuation modelling process. The
historical data for the study area are taken from the National Geophysical Data Centre (NGDC)
where historical observations are available. However, this data base is incomplete. An examination
showed that many known tsunami events are not mentioned. A further limiting factor of this
approach is the consideration of worldwide events in the analysis and not only the tsunami events
which affected Indonesia. The intention of an analysis of historical tsunami events is to find out
which magnitude led to what situation at the coast in the historical context, regarding horizontal
and vertical inundation and the certain occurrence probabilities. However, this approach does not
consider the different regional conditions for tsunami generation and propagation. A decisive point
76
is the bathymetry of the seafloor because the speed, and ultimately the direction, of the tsunami
are controlled by the depth of water. Therefore the tsunami impact on land strongly depends on
particular regional circumstances (GTZ- IS 2007).
Numerical modelling of tsunami inundation on land delivers information on intensities of possible
tsunami inundation events for a certain location. For the study area only one inundation scenario
could be used for the evacuation modelling process because the results of the multi- scenario
approach did not provide safe areas within the study area (cp. Chapter 4.2.1). According to a main
objective of this thesis, giving a statement about the endangerment of the study area, using a
multi- scenario approach would provide a comprehensive data basis. An first efficient way to
establish tsunami hazard zones using multiple scenario results is by counting on each point on the
land if this location is inundated or not. Incorporating all scenarios - which should be repre-
sentative for all possible tsunami events in the considered region - leads to a frequency
distribution of each location on land being exposed to a tsunami or not. Although not directly
indicating an expected intensity, the resulting information may serve decision makers to
characterise zones by the fact of frequency being exposed. This means that information are
provided based on the frequency being hit rather than on expected intensities for a single case
event (STRUNZ ET AL. 2008A).
In a further step of analysis the gained impact zones can be modified by relating the impact areas
to tsunami warning categories which are derived from tsunami wave height at the coast. Tsunami
Early Warning Systems currently relate the level of warning (e.g. minor, major warning) to the ex-
pected wave height at the coastline. By deriving the inundation zone for each warning level (wave
height), one can relate the warning levels to hazard impact zones and safe areas within the hazard
map. This enables to plan and prepare disaster management and evacuation planning strategies
for different tsunami intensity classes (STRUNZ ET AL. 2008A). As a further step, guidance for evacu-
ation can be related to the warning levels. Different warning levels and hence different estimated
tsunami intensities lead to differentiated evacuation demands. For example, in case of an “Ad-
visory” warning with a related wave height from 0 to 0.5m, it is not necessary to evacuate a whole
city or a bigger area, because only the areas near the coast are affected. A simple guidance like
“move away from beaches and rivers” (GTZ-IS 2008) is sufficient in this case. Apart from the
general necessary relation between warning level and evacuation guidance, it is of prime im-
portance that people at risk have to know exactly how to react on which information. There-
fore standardized evacuation guidelines for each case are necessary to implement fixed evacuation
procedures for the population and for the local decision makers. In case of a warning the popu-
lation have to know if, depending on the warning level, they are located in potential inundation
area and how they have to cope with the situation. The responsible decision makers need clear
and easy understandable information of the correlation between a special warning level, the
related tsunami impact zone and the related evacuation guidance. Figure 54 below clarifies this
combination.
77
Figure 54 Useful combination between hazard impact zones, warning levels and evacuation guidance (Source: own illustration - Data basis: GTZ- IS (2008), modified)
Results obtained by numerical modelling possess a certain degree of uncertainty (due to
incomplete knowledge on model parameters, input data assumptions, spatial resolution of input
data as e.g. bathymetry and consequent error propagation) (STRUNZ ET AL. 2008A). These
uncertainties in modelling results have to be considered when establishing the tsunami affected
area for further analysis regarding evacuation planning. This can be done either by incorporating
model uncertainty assessment results (e.g. uncertainty margins for each respective model output)
or as a first approximation to assign a buffer value to simulation results based on expert
knowledge or by validating on historical event data.
These described technical uncertainties refer to an important fact which concerns the whole
discussion about tsunami hazard assessment and evacuation planning. Evacuation procedures
based on hazard maps need to be disseminated to the local population. It should not be ignored
that the underlying data for all procedures are only based on scenarios, estimations of experts or
historical experiences. However, the exact process of a tsunami event is not predictable and
therefore all recommended evacuation procedures, especially calculated inundation and safe areas,
have an uncertainty factor that have to be considered. Evacuation plans have to be designed in a
way that allows a flexible adaptation to changing situations. Scenario- based hazard information
are only a decision support which have to be used for a fast calculation of a situation in order to
78
provide an easy and standardized evacuation guidance for an expected tsunami event. The
population have to be sensitized to the dynamic process of tsunami events as well as to the
appropriate evacuation procedures which have to be handled in case of a changing warning
message.
5.3 Population modelling
The general importance of detailed population data for evacuation planning on local level and
population disaggregation methods are already described in Chapter 3.1.3 and Chapter 3.3. An
aim of this research was to develop an accurate representation of population distribution in the
study area, estimating daytime versus nighttime fluctuations in the event of a tsunami. The
implementation of a micro scale map of population density in the daytime and nighttime provides a
realistic snapshot of who is on the ground when the tsunami event occurs, highlighting higher risk
areas during response and evacuation efforts (SLEETER AND WOOD 2006).
Many methods to disaggregate population have been practiced in the GIS and Remote Sensing
fields. TAUBENBÖCK (2007) raise the crucial questions, (1) how many people live where (within an
urban agglomeration) and (2) how does population concentration change spatially within one day.
With an object- oriented classification methodology, structural criteria like built- up density,
building height and land use can be derived and, based on the hypothesis that homogeneous
structural urban sectors show homogeneous demographic characteristics, the local population
distribution can be indirectly derived. Hence, the first question can be answered. Considering
structural characteristics of housing units, like particular roof types, building size and location,
different land use classes, like residential and commercial areas can be assessed. Assuming that
commercial areas are predominantly populated during normal working hours and residential areas
during the night, a weighting factor can be used to consider spatial population dynamics within
one day. This approach was not goal- oriented for this research because of more detailed
demands concerning spatial behaviour patterns of the extremely mixed tourist and local population
in the study area. Furthermore the process of an object- orientated urban land cover classification
is very time- consuming and could not be conducted in this research.
The division of the study area into functional urban sectors is an important prerequisite to visualize
population movements during day and night regarding their activities in the particular sectors in a
certain time flow. The main objective was to sign out general “activity- areas” which mainly show
one certain function. Therefore it is not necessary to know each building function but only to
conceive a particular character of an urban sector. Thereby the main challenge was to quantify the
tourist population. While the local population is moving predominantly between residential areas
and commercial areas, the tourist population show very heterogeneous moving patterns. Tourist
movements are the spatial changes of activity locations of tourists and are motivated and affected
by the attractions and activities available at the destination (LAU AND MCKERCHER 2006). Another
very dynamic population group which play an important role in the study area are the daily
79
commuters. The area of Kuta is one of the most touristic places in Bali and therefore offers good
opportunities to earn money, mainly in the informal sector within the tourist centres. This fact
leads to massive population movements during the day and have to be considered.
This mixture of local and tourist population in a relative small area was reflected in the effort of a
functional pre- zonating on the basis of a high- resolution satellite image. The only functional
sector which showed very clear structures were the residential areas for the tourists , because of
the regular formation of Hotels and Bungalows. Apart from some small commercial units, the area
could not be assigned to homogeneous functions. A field work was essential to assign functional
urban sectors to the particular population group and to quantify the population movements within
one day. As presented in Figure 22, the study area shows a very complex structure of functional
units and seems to be very unclear at first view. But regarding the assigned quantifications of the
particular sectors, it becomes apparent that the differences between the day and night population
are considerable and play an essential role in the process of evacuation planning. It is of great
importance where concentration spots of people are at a particular time to meet the requirements
of evacuation planning at local level. In Chapter 5.2, the need of standardized evacuation
guidelines for different tsunami warning levels is described. Related evacuation plans identify
affected and safe areas, evacuation routes and options for vertical evacuation, but the real
evacuation process stands and falls with the availability of suitable population data. If the real
population distribution in an affected area differs widely from the estimated distribution as the
basis for an evacuation plan, the evacuation process during a tsunami event will not achieve the
planning objectives. Displayed evacuation routes could be blocked because they are overcrowded
with people, capacities of evacuation shelters could be insufficient because wrong evacuees
numbers are calculated and the accessibility times from source points to the nearest evacuation
shelter could be extended by a reduced evacuation speed. In return, a wrong estimated population
distribution could lead to a neglect of resources like little frequented evacuation routes and half-
empty evacuation shelters.
The results of the two applied disaggregation methods in Chapter 4.3 demonstrate the wide
variations between different approaches both regarding the general distribution and regarding the
day and night values. Looking at Method A, it becomes apparent that a simple weighting factor for
the differentiation of a day and night distribution does not meet the demands for local evacuation
planning within urban agglomerations. Furthermore the disaggregation shows too spacious results
to supply information regarding evacuation modelling. On the one hand, this method is generally
more concentrated to weighting factors concerning the presence of people in areas of different
types of land use. On the other hand this static approach does not consider dynamic processes like
touristic behaviour patterns and commuter streams.
Regarding the demands for evacuation planning on a local level on the one site and the different
approaches for population distribution in combination with the appropriate efforts on the other
site, it can be stated that for this thesis the development of a new method for disaggregating
population was necessary. Neither a bottom- up approach (building- based population data
80
collected by fieldwork; see Table 10) nor a known top- down approach (population distribution
according to urban structural characteristics derived from remote sensing data, see Table 10) was
a suitable method to achieve positive results as a basis for evacuation modelling. The conducted
field work was a very time- consuming process but indispensable to display the dynamic
population distributions in a touristic urban agglomeration within one day. The gained outcomes
provided a stabile base for the subsequent accessibility and evacuation modelling.
Regarding the important objective of population modelling, highlighting higher risk areas during
response and evacuation efforts, a further parameter could be included the study: the different
vulnerability of social groups. In order to estimate the current susceptibility and coping capacities
of different social groups and households to tsunami hazard, a UNU-EHS study (BIRKMANN ET AL.
2007) tested different indicators in order to explain the revealed vulnerability, and also to estimate
the present vulnerability and coping capacities. The following indicators were selected and tested,
in terms of their relevance to visualise different revealed vulnerabilities:
- Amount of young and elderly people in the total population
- Gender
- Income
- Employment
- Land ownership
- Social networks and membership of organisations
- Loans and savings
- Potential recovery time (index)
The analysis of the distribution of the dead and missing people after a tsunami event according to
the different social indicators initially shows that in two of the most affected cities (Galle and
Batticaloa) gender played an important role with regard to the dead and missing, while the
number of females is significantly higher than for males. Many reasons have been given for this
from the observations of rescue teams and comments by survivors:
(1) more women were at home looking after routine domestic activities
(2) when the tsunami hit, women caught by the waves got their dresses (sarees) and hair
entangled with objects in their path (fences, trees, etc) causing death
(3) women were trying to look after children and were delayed at home after being warned, while
others were fleeing
(4) women tried to lock up houses and collect precious items, which delayed them moving out
The analysis also shows that young, elderly and disabled people are highly vulnerable mainly
because of limited mobility, strength and balance. Therefore it is important to target these most
vulnerable groups first and as a priority in evacuation situations (BIRKMANN ET AL. 2007).
Due to the lack of time, for the study area vulnerable groups are only defined by allocating
building functions to certain social groups: (1) Schools are localized to show a great number of
81
children during a particular time and (2) the location of hospitals represent a great number of
people who need special help during an evacuation. Using questionnaires as a data- gathering
tool, the vulnerability of different social groups could be highlighted much better and implemented
into an evacuation modelling process.
Regarding the gained results in Chapter 4.4, the advantage of Method B over Method A is already
described since the results of the field work provide a detailed insight into the population
movements of the study area. Looking at Figure 22, a functional dichotomy of the area is in
evident. The tourist population is located along the coast while the local inhabitants are living
further inland. The majority of the population are local people; the tourists come up to about one-
fifth of the total population in the study area. This leads to the following statements regarding
evacuation planning:
(1) The tourists are the most vulnerable population group to tsunami impact in the study area at
any time of the day. Due to the functional mix of residential and business sectors near the coast
(cp. Figure 22), the whole area is highly frequented both during the day and the night.
(2) The majority of the population in the study area live further inland but even so within the
estimated tsunami hazard zone (cp. Figure 31, 32 and 28).
(3) The most vulnerable areas are the tourist business sectors during the day because of the short
distance to the coast and because local inhabitants and commuters stay in these areas during the
working hours.
However, these statements are only based on the developed population distribution concept used
in Method B. A validation of this concept is required, since the weighting values for the functional
sectors are only based on field observation and expert interviews. A literature research was not
successful since no comparable functional zonation for coastal areas could be found.
5.4 Accessibility modelling
Accessibility modelling is the core of the evacuation modelling process while it provides the
evaluated surface for all evacuation procedures. While evacuation plans are commonly created by
choosing main roads as suggested evacuation routes, the most important additional benefit of this
approach is the consideration of alternative evacuation routes beside the road network. In a real
evacuation situation the road network will be the most frequented route, but if other land use
surfaces seem to provide a shorter path to the next evacuation shelter these routes will also be
used. While different types of evacuation infrastructure have different characteristics, it is not
enough to consider only the distance values. Instead, a measure of travel cost (Cost Weighted
Distance) is preferable. A further benefit of this approach is that the cost is measured as travel
time and the cost to travel across different land use classes is estimated considering speed
impedance values. For the study area impedance values were assigned only considering the
surface of each passable land use class. Based on a surface with no impedance, for the road
82
network the width and the traffic volume are the speed limiting factors. The selection of the
certain values for each land use class was based on a study from the Asian Disaster Preparedness
Centre (ADPC) (ADPC 2007B). To test the effect on the evacuation time for a defined distance, the
values for each land use class were changed a few times and assigned according to an own
estimation. A possible validation of these estimations could be empirical investigations of people
movements over different land surfaces or a further literature research.
The CWD approach offers the possibility to use a lot more parameter improving the simulation of
evacuee’s movement. Slope of terrain is one of the factors that most influence the evacuation
speed; walking or running on a steep area would be more expensive in terms of physical exertion
than doing the same on a flat zone. For this reason it is an important input in evaluation of Cost
Weighted Distance and can easily combined with the speed conserving values for the different
land use classes. For the study area slope values are not considered because the area is very flat
and therefore an appropriate calculation would not be useful. As described in Chapter 3.4.2, the
evacuation speed from a source point to the nearest evacuation shelter is defined as the average
speed of evacuation (1,2m/sec) combined with the reclassified (speed limited) land use value and,
if meaningful, the reclassified (speed limited) slope value. Regarding the average speed of
evacuation, it is emanated that one person can walk 1,2m/sec in average (ADPC 2007A). Other
speed values could found in literature during the data research. Table 26 shows some estimations
from other data sources.
Walking condition Average walking speed (m/s) Source
A person pushing a perambulator 1.07 Institute for Fire safety & Disaster preparedness (SUGIMOTO ET AL. 2003)
A person with a child 1.02 Institute for Fire safety & Disaster preparedness (SUGIMOTO ET AL. 2003)
An independent walking elderly person
0.948 Institute for Fire safety & Disaster preparedness (SUGIMOTO ET AL. 2003)
One person (not specified) 1.38 KLÜPFEL ET AL. 2004
Table 26 Average evacuation speed values from literature (Source: own composition)
83
An important aspect during an evacuation is the dependence of the walking speed on the group
size, which is not considered in this analysis. The Traffgo GmbH (KLÜPFEL ET AL. 2004) tested this
dependence with different group sizes. The results showed that different group sizes greatly
influence the particular velocity, while bigger groups moved slower than smaller ones. This
parameter could also be considered in the analyses by establishing a relationship between the
population density per raster cell and the related time necessary to walk through the cell. A cell
grid with the gained velocities could replace the fix average speed of 1,2m/sec. Figure 55 below
shows this relationship.
Figure 55 Potential improvement of the CWD method (Source: own illustration - Data basis: CWD approach)
Regarding the defined evacuation shelters as the basis for accessibility modelling, it has been
already referred to the pre- operating character of this study, while shelter points are only based
on a scenario and own assessments. All results of this study are only meant to serve as spatial
information and recommendations for official evacuation planning. The official definition of
evacuation shelters in the study area can only be a political decision and is therefore a very
sensitive topic. In the study area no evacuation shelters were defined officially during the research
process. Due to the fact that the topic of evacuation planning is just at the outset in the province
of Bali, the related official decision structures are currently not clearly defined.
Local decision makers were sensitized to this topic during the field work by workshops on
evacuation planning supervised by GTZ- IS. Thereby central discussion points on the part of the
local representatives were the nescience about potential affected areas in case of a tsunami as
well as questions to location and requirements of evacuation shelter buildings. These uncertainties
and doubts are a result of the insistent note that, on the one side, all assigned risk areas are only
84
based on numerical inundation models or historical data, but, on the other side, that it is of great
importance to define fix buildings which meet the requirements for evacuation shelters. A more
specific problem are the different structures of ownership in Bali. Due to the wide touristic
infrastructure, hotels are the most important potential evacuation shelters. The hotel industry in
Bali is predominantly private sector and the public influence capability is limited. For economical
reasons, the owners are often not interested to broach the issue of tsunami too much, to avoid a
relation between the real touristic product and a potential tsunami hazard. However, the tsunami
issue is more and more in people’s mind and the issue tsunami safety plays more and more an
important role, especially in coastal areas. Hence, hotels are increasingly interested to mark their
facilities as “tsunami safe”. Thereby a central point will be, how far the hotel owners are poised to
accommodate not only their own hotel guests but also evacuees from outside in case of a tsunami
event. These discussions have to be led between the local decision makers for evacuation planning
and the local hotel industry.
The concept of vertical evacuation and ESB should also be enforced and supported with the
implementation of building codes and regulation for land use planning and rehabilitation. The
implementation will rely much on local government’s role rather than national-wide authority
(BUDIARJO 2006). On the one hand, known building functions that can be assigned as evacuation
shelter building, like mosques, schools, offices or shopping centres can be implemented in the
facility planning process regarding general requirements for evacuation shelters. On the other
hand, in case of vertical evacuation, people actually can also evacuate themselves to the upper
floor in their own houses (if the houses fulfil basic ESB requirements). In this case, the
construction strength is questionable because it may not be able to withstand tsunami waves.
Therefore the current inadequate enforcement of building regulation and construction quality
control has to be improved.
5.5 Evacuation modelling
The great challenge of the evacuation modelling is to combine the results of the accessibility
modelling with the disaggregated population data. Thereby the temporal dynamic of an evacuation
process has to be represented as best as possible. The relevant information which are needed to
calculate people’s ability to evacuate, presented in Figure 17, were the basis for this process and
will be discussed regarding the gained results in Chapter 4.5.
First of all, the calculation results for the study area can only be evaluated with some limitations.
(1) The evacuation modelling is based on different population distributions considering census data
from 2000, estimated tourist and commuter data (cp. Chapter 3.3) and two population
disaggregation concepts which are not validated.
(2) The identified hazard impact zones are only based on one potential inundation scenario. A
multi- scenario approach would provide a better data base (cp. Chapter 3.2).
85
(3) The identified evacuation shelters (safe areas and ESB) are not approved officially and the
shelter capacities are partly estimated (cp. Chapter 4.1 and 4.2.1).
(4) The accessibility modelling and accordingly the evacuation time values are based on an
estimated cost surface (cp. Chapter 3.4.2).
The modelling results deliver very important information. The different values of the shelter
accessibility for day and night population as presented in Chapter 4.5.1 are a result of the different
population distributions. Thereby a crucial information is the relative good accessibility for the
major part of the population, independent from the based population data. While the location of
safe areas for horizontal evacuation is based on an inundation scenario and therefore
predetermined, it is possible to improve the evacuation of a large area within a short time (15min
in the study area) by a reasonable distribution of potential evacuation shelter buildings. This
information is only based on the Cost Weighted Distance (CWD) approach as estimated parameter.
As discussed in Chapter 5.3, this approach can be improved by considering further parameters like
group size or by an evaluation of the defined speed values for different land use classes. The
method for the shelter accessibility can be an important and transferable spatial information for
official evacuation planning, since areas with bad evacuation ability are in evidence and all
necessary action can be initiated. An indispensable basis for such information is an appropriate
population distribution to ensure a correct identification of required evacuation shelters. The maps
presented in Chapter 4.5.1 only show the shelter accessibility for each shelter catchment area on
the basis of different population distributions. A decisive additional information is, if the based
population data are sufficient to describe the evacuation situation of areas with different
population densities. In Figure 56 and 57 the results of the shelter accessibility for the day
population calculated with Method A and B are contrasted with the based population data. Looking
at Method B, detailed statements regarding the emptying of areas with different population
densities are possible. It is evident that very densely populated areas in the centre of the study
area show good accessibility values, while a large number of people near the coast is not able to
reach the nearest evacuation shelter within 15min.
86
Figure 56 Analysis of the evacuation shelter accessibility within 15 min regarding the population distribution (Method B) in the study area (Source: own illustration - Data basis: CWD surface and population data calculated with Method B; Map basis: Figure 31 and 46)
Figure 57 Analysis of the evacuation shelter accessibility within 15 min regarding the population distribution (Method A) in the study area (Source: own illustration - Data basis: CWD surface and population data calculated with Method A; Map basis: Figure 29 and 48)
Comparing the population distribution of Method A with the related shelter accessibility calculation,
no specific statements regarding population hotspots and their ability to access a certain shelter
are possible, because the population map is too coarse.
87
The two sample areas show the number of population disaggregation units calculated with Method
A and B and the related number of population per unit. The different population values are a result
of the disaggregation approaches and are already discussed in Chapter 5.3. Important for the
evacuation modelling is also the spatial resolution of the population distribution. Sample area A is
subdivided into 17 sectors, sample area B into 43 sectors. In case of a tsunami event, the
temporal evacuation process within sample area B can be monitored more detailed since the
allocation of people to the evacuation shelters can be planned exactly ( particularly regarding
shelter capacity) and the evacuation situation of a spatial sector at a particular time is easy to
determine. The highlightning of sensitive areas with a weak potential to access evacuation shelters
is an important spatial information for evacuation planning, but can only come into value by a
detailed based population distribution.
Further important information are the locations of central evacuation bottlenecks within the study
area. That means, (1) where converges a high population density and potentially high frequented
evacuation routes and (2) where are coastal areas with a bad ability to evacuate.
In Figure 58, the main evacuation bottlenecks in the study area are identified and will be discussed
in the following regarding the functional sectors in Figure 59.
Figure 58 (left) / 59 (right) Left: Evacuation bottlenecks in the study area (Source: own illustration - Data basis: population distribution calculated with Method B and the evaluated road network; Map basis: Figure 31); Right: Functional urban sectors in the study area (Source: own illustration - Data basis: own data collection; Map basis: Quickbird satellite image)
88
Bottleneck 1 shows a mixed commercial district for the local population and the tourists which is
densely populated both during the day and the night. The area is located at an important traffic
junction in the study area and therefore a main evacuation bottleneck. People within the
commercial area and also people from the beach will use these roads to evacuate and therefore a
traffic jam and a slow evacuation speed can be expected in case of an evacuation. Bottleneck 2
shows a touristic commercial area near the coast which is densely populated during the day. The
main road from the South to the centre is an important communication road to the main hotels
districts and beach areas in the study area. In case of an evacuation, mainly tourists will use this
road. Bottleneck 3 and 4 show residential and commercial areas of the local population. Both
during the day and the night these areas show the most activity of the local population. Hence,
the main roads along these areas will be highly frequented during an evacuation. Bottlenecks 5
and 6 exemplarily show two beach areas in the South of the study area. The green routes identify
the fastest evacuation route to the nearest evacuation shelter. Figure 40 reveal that many areas
along the coast show high evacuation times and respectively the worst evacuation shelter
accessibility in the study area.
These statements are only based on estimated values (speed impedance, population density) and
are only useful to sum up the potential situation during an evacuation and to get an overview of
potential bottlenecks.
The evacuation modelling was conducted to give a decision support for evacuation in case of a
tsunami event. Thereby a decisive parameter is the temporal and spatial dynamic of an evacuation
process (DE SILVA AND EGLESE 2000). While this study is based on a static approach, dynamic
components were tried to compensate with static parameters. The based cost surface can simulate
the emptying of an evacuation area by stating values which show how many people are able to
evacuate within a certain time and within a defined area. Considering day and night population
distribution, changing situations during one day can also be showed. Figure 17, presented in
Chapter 3.5, shows a further important factor for evacuation modelling: the need for additional
evacuation shelters in certain areas. This point is based on the results of the previous accessibility
and evacuation modelling where it can be stated if there is a need or not. However, more detailed
information are required. The following questions have to be answered to deal with this problem:
(1) How much time does it take until an evacuation shelter has reached the maximum capacity?
(2) How many people have to be evacuated to another shelter in the remaining time?
(3) Are additional capacities available in surrounding shelters?
(4) Is there a need for an identification of additional evacuation shelters and what capacity is
required?
These questions can also be answered with a static approach. The evacuation modelling in
Chapter 4.5 calculates the shelter accessibility in the study area within 15 minutes. By an
application of the modelling approach with different time frames (e.g. 5, 10, 15, 20 minutes) it is
possible to calculate the required time until an evacuation shelter has reached the maximum
capacity. The number of remaining people which have to be evacuated to another shelter can be
89
calculated by the difference of the people in the catchment areas which are potentially able to
reach the nearest shelter within a given time, and the people who already reached the assigned
shelter. Table 27 provide a rough exemplarily calculation for additional required evacuation shel-
ters in the study area based on population data gained by Method B. Column two and three show
the number of people who are able to evacuate to the nearest shelter within 15 minutes (cp. Table 24). For some shelter catchment areas the capacity is not sufficient. By comparing the maximum
number of people in each catchment area with the shelter capacities, the required additional
capacity can be easily calculated. The gained results also provide information about the required
building type. While for catchment area number five a small building with a low capacity is
sufficient (e.g. a market building), the areas four, six and seven need one or more bigger buildings
which meet the general requirements for ESB (e.g. hotel, sport hall).
Basin_No. Pop_15min Shelter capacity
(number of people)
Additional shelter (Yes/No)
Required shelter capacity
(max. number of people)
Method B_day Method B_night
1 531 409 776 NO ----
2 1467 842 2000 NO ----
3 577 849 1125 NO ----
4 1891 1059 226 YES 1665
5 1067 1049 824 YES 243
6 2988 3409 526 YES 2883
7 2061 2434 214 YES 2218
8 141 99 151 NO ----
12 2685 3077 97375 NO ----
21 1721 1753 77320 NO ----
Table 27 Exemplarily calculation for additional required evacuation shelters in the study area (Source: own composition – Data basis: population distribution concept (cp. Fig. 13)
Regarding point (3), available capacities of surrounding shelters can be calculated simultaneously
in the same way and can be compared with the number of people from other catchment areas
which still have to be evacuated. In case there are still evacuees who are not able to evacuate
because of insufficient capacities, additional shelters have to be identified.
90
The main disadvantage of the static approach in this research is that the evacuation modelling
cannot be showed in a dynamic process. To improve the modelling accordingly, dynamic models
dedicated to analyzing and displaying interaction of flow data are required. The Flowmap software
(BUDIARJO 2006) can be mentioned as a suitable tool for dynamic processes regarding evacuation
modelling by providing useful measures for accessibility analyses. The catchment area analyses feature provides a possibility to allocate disaggregated population units to the nearest evacuation
shelter. Flowmap allows two bounds to be set: a destination (evacuation shelter) can have a
maximum capacity, or destination can have a maximum reach (distance, travel time) (ZWAN ET AL.
2005). In the former, if the maximum capacity is reached, the destination is not taken into account
any longer in the remainder of the allocation procedure. While in the latter, origins that fall outside
this reach cannot be allocated to this destination. The maximum reach must be set and is equal for
all destinations. The result from catchment area analysis include: the number of allocated
population to a destination, the catchment area of each destination, the distance or travel time
from each origin (cell) to its destination, and the remaining non-allocated population number can
be obtained. The Service location planning modelling is a further Flowmap tool that can find the
location of new additional evacuation shelters required to population that is not covered by the
existing potential shelters. The tool attempts to allocate additional evacuation shelters particularly
at the location which contains high population number (high density area). The option is useful
because the shelters should be located at high-density areas where the population is high rather
than at the low-density areas. To combine or incorporate the existing potential shelters with the
proposed additional ones, catchment area analysis is conducted to get the location, shelter
catchment area and capacity of each allocated shelter as well as evacuees’ travel time from each
cell to its allocated evacuation shelter.
5.6 General discussion
In the following, the particular results of the previous process steps will be discussed in the overall
context considering the gained findings for the study area and the transferability of the used
approach.
Due to the good initial data basis for Bali, the possibility to conduct a detailed pre- operating
research in the study area was given. Available numerical inundation models, land use and street
data as well as rough population data were important input data for a first estimation of the
tsunami risk for the study area. But for all that, the high demands on evacuation planning
necessitate a comprehensive field work. The critical points in very dense urban areas like the study
area are detailed population data integrating tourists and commuters as well as applicable
evacuation shelters. Looking at the results of the accessibility modelling, it becomes apparent that
vertical evacuation plays an important role in urban evacuation planning (BUDIARJO 2006).
Especially the hotel sector has to be integrated in the planning process due to the huge amount of
adequate buildings in the study area. Although the modelling is based on a subject to the shelter
selection, it becomes apparent that it is necessary to select an applicable spatial structure of these
91
buildings to ensure a complete coverage of the area. Important criteria are the consideration of
population hotspots and evacuation bottlenecks, because in case of an evacuation many different-
sized population streams using the few evacuation routes, will evacuate to the next evacuation
shelter. Defining population hotspots also requires a dynamic regard of the affected population to
respond to changing situations. The disaggregation of the available census data to identify spatial
and temporal population movements within one day is an adequate methodology to meet these
demands. Thereby, the functional zonation of the study areas shaped up as a suitable approach.
Apart from the general evacuation infrastructure, the evacuation time respectively the number of
people which is able to evacuate on the basis of an estimated cost surface is a main concern in
this research. The analyses show that most of the affected people can reach an evacuation shelter
within an estimated time span of 40min (cp. Figure 40), but regarding the assumed values a
subsequent sensitivity analyses is necessary to validate these results. For models involving many
input variables this CWD method is an essential ingredient of model building and quality
assurance. Thereby the influence of different input factors to certain output quantities will be
analysed. Relating to the study approach, the influence of the speed impedance factor for the
evacuation time map based on different land use classes as well as further relevant input
parameter like slope or mass evacuation values can be analyzed regarding the particular
evacuation time.
Regarding the transferability of the research results, the concept of Cost Weighted Distance (CWD)
was proved to be a suitable approach mainly by considering alternative evacuation routes beside
roads. Their important function in the study area, as interface between roads as the main
evacuation routes become evident. Especially for evacuation planning in high- populated urban
areas the concept is very useful to relieve the high frequented road network. The technical
implementation of the concept on a small scale is without any difficulty and well transferable to
other coastal areas. However, the point- based access to evacuation shelters is a critical factor.
While for the vertical evacuation to ESB’s this approach is correct, the punctual accessibility to safe
areas outside inundation zones is a limitating factor. For the study area, wide streets as the most
important evacuation bottlenecks were chosen as access points. For an analysis on a small scale
this method is suitable to simulate an evacuation situation, but on a broad- scale level some
problems can occur. To ensure a realistic accessibility, the access points have to be set very dense
and regular along the border of an inundation zone. However, this entails a big manual effort and
processing power. A further profit of this concept is the raster- based approach which allows an
adaptation to different levels of detail. While for the small study area a 1m cell size was chosen to
ensure a detail analyses, this resolution is not applicable for larger areas. By considering the
defined cell size in the formula for the accessibility modelling, different scales are possible.
A crucial point of this research is to keep in mind the link to the local administration and decision
makers in the study area. According to the research objectives of this thesis, the requirements for
evacuation planning are defined, the potential tsunami threat in the study area is analysed and by
the development of a spatial information system, a useful decision support is provided. Based on
92
the gained expertises during this research it is possible to define some important stakeholders
which should be incorporated in the research process. Each mandated institution ideally has a
Standard Operating Procedure (SOP) as the guidelines and the procedures to deliver its services.
(1) A hazard assessment has to provide scientifically information products as well as simplified
maps for the local communities. Therefore it is of great importance that the development
integrates knowledge from research or technical institutions, e.g. LIPI, BAPPEDA, BMG and
Bakosurtanal.
(2) Population modelling is a crucial topic in evacuation planning as described in detail in the
previous chapters. BPS can provide the necessary base data for further analysis or can conduct
field surveys to gain more detailed population data. The neighbourhood communities play an
important role in the Indonesian society and should be involved to get more detailed information
on the socio- economic structure of communities.
(3) For the identification of suitable evacuation routes it is essential to know the location of
population hotspots, evacuation bottlenecks and have information about the traffic volume in the
course of the day. Community leaders or local NGO’s (Non- governmental organisations) can help
to identify these parameters. PMI has much experience in emergency planning and should also be
involved in the planning process.
(4) The assignment of evacuation shelters is of great importance. The identification process of
suitable shelters should involve structural engineers, TNI (Armed forces of Indonesia), research
institutions as well as the local and regional administration.
(5) The identification of vulnerable groups (elder people, children, women and handicapped
people) which need special attention during an evacuation is essential. Specific evacuation
measures are mainly related to transport facilities. TNI, police and maybe private sectors will be
the responsible institutions for transportation and have to be involved in the planning process.
(6) Traffic control and security is also task of the police.
(7) Evacuation procedures need to be disseminated to the local population. Information materials,
maps and signs will help to inform general public about evacuation procedures. This process
requires the cooperation of local governments, community leaders, private sector, local academic
institutions and NGO’s.
93
6 Conclusion & Recommendations
Conclusion
Generally, the thesis research can satisfy the research objectives and answer the research
questions addressed in Table 2. The research aims to develop and apply a spatial information
system supporting tsunami evacuation planning.
In relation to each research objective, the research can be concluded as follows:
Definition of requirements for evacuation planning
The requirements for evacuation planning are derived from the related evacuation planning
objectives, determining exactly what tasks need to be carried out once the decision to evacuate
has been made. A community risk assessment can be seen as the starting point for evacuation
planning by providing the basis for an appropriate evacuation strategy (EMERGENY MANAGEMENT
AUSTRALIA 2005). To define the degree of tsunami impact on land, a hazard assessment provides
information about the nature, location, intensity and probability of a tsunami threat. Numerical
modelling of ocean tsunami wave propagation delivers estimated arrival times at the coastline,
time duration of the tsunami event as well as inundation and wave heights above land surface.
Using these information tsunami impact zones along the coast can be estimated. Historical records
of tsunami impact observations are also important sources to gain information on probabilities of
tsunami events with a certain magnitude and to derive relationships of hazard impact zones on
land when tsunami modelling results are not available (POST ET AL. 2007).
In case of a tsunami the decision whether a region should be evacuated requires information on
the spatial distribution of exposed population and critical facilities (schools, hospitals, etc.) in order
to provide information about local evacuation “hotspots” and “uncritical” areas. Due to the fact
that population data on a smaller scale better satisfy the demands for local evacuation planning,
methods to disaggregate population data should be applied while providing more detailed
information. Regarding different concentrations of people over the day, a dynamic parameter
considering temporal and spatial population movements can improve the information quality.
For evacuation two methods are generally available: horizontal evacuation, moving people to more
distant locations or higher ground outside the inundation area (safe areas), or vertical evacuation,
moving people to higher floors in buildings (ESB). Safe areas can be derived from numerical
inundation models. Adequate accommodation criteria like sufficient capacity, adequate ground
surface and low slope values have to be determined in addition. Robust multi- storey buildings
assigned as the destination in vertical evacuation have to meet specific requirements of structure,
evacuation floor, function, design, capacity, accessibility and security (BUDIARJO 2006). Evacuation
routes have to be defined considering main streets and other passable land surfaces like small
vegetation or agricultural crop land. Roadways and intersections that could be evacuation
bottlenecks should be highlighted to show main emptying structures.
94
Development of a transferable and applicable spatial information system for tsunami
evacuation planning
First of all, the developed spatial information system should meet the technical demands to
implement the most common input data required for evacuation planning. Big amounts of input
data, like satellite images and spatial base data have to be processed. Shapefiles and raster data
are the main used data types because of the area- wide availability. A spatial information system
based on ArcGIS seemed to be the most stabile solution for this research. Special software
packages for evacuation modelling were tested but not used due to problems with data types and
size. The methodical composition of the developed information tool is clearly structured and
documented both regarding the technical process and the required input data. The methodical
approach leaves room for further improvement like additional input data and technical extensions
and can easily adapted to changing demands in other tsunami- prone areas.
Giving a statement about the endangerment of the study area
Substantiated statements about the endangerment of an area are based on different components.
As a process in risk assessment it is generally agreed that it includes (1) the determination of the
potentially endangered coastal regions by conducting a tsunami hazard assessment, (2) the
estimation of the vulnerability of the people and the environment in the affected areas and (3) the
identification of capacities and resources available to address or manage a tsunami threat
(coping). (TAUBENBÖCK ET AL. 2008)
Based on the results of one potential inundation scenario, the affected coastal regions of the study
area are mapped, excluding the various spatial intensity and probability of the tsunami occurrence.
The scenario is based on the current knowledge on the possible ocean bottom deformations due to
seismic activity in the Sunda trench region and was calculated for a moment magnitude of 8.5.
To provide a good insight into the spatial distribution of the population at risk, available input data
were disaggregated to smaller sectors considering day and nighttime fluctuations. The
identification of facilities featuring particularly endangered people improves the quality of
information. Coping is the mechanism by which people are able to reduce effectively the impacts
of the disaster and facilitate their lives as far as possible (BIRKMANN ET AL. 2007). The capability to
evacuate, regarding physical infrastructure and available evacuation time, is determined in this
research by the accessibility and evacuation modelling results.
Giving a decision support for evacuation in case of a tsunami event
The decision to evacuate highly depends on the severity of the expected effects of the disaster on
the exposed population and the ability to evacuate. Planning for an emergency evacuation
therefore involves addressing both behavioural and logistical issues that greatly influence the
evacuation operation (DE SILVA AND EGLESE 2000).
95
Logistical problems in evacuation can arise during the warning process, when transporting and
providing accommodation for large numbers of evacuees or while administering protective
measures. The behavioural issues encompass the complications that arise due to the process of
mass evacuation while survival instincts overcome rational knowledge. An evacuation can
therefore end up to a chaotic situation which the emergency planners must be prepared to
manage in order to avoid danger, confusion and death. Local responsibilities can therefore greatly
benefit from a GIS- based spatial information tool which can provide products to examine the
effects of different assumptions or different plans.
Local decision makers have to know the locations of hotspots which are of vital importance during
an evacuation. This can be areas with high population density at a particular time, critical facilities
(schools, hospitals) in affected areas or evacuation bottlenecks which will be strongly frequented
during an evacuation. Changing population distributions during day and night lead to different
evacuation demands. A population distribution concept considering day and night values which
was developed in this thesis (see Chapter 3.3) can provide useful information for temporally
differentiated decision processes. Facilities featuring particularly endangered people were identified
during the field work (see Chapter 4.1) and therewith population groups which need special help
during an evacuation are highlighted. Central bottlenecks of the study area are illustrated in Figure 58 to demonstrate the main frequented evacuation routes.
Both the emergency planners and the exposed population have to know the fastest evacuation
route to the nearest evacuation shelter from every point in the affected area and the estimated
time to go there. The CWD concept (see Chapter 3.4) provides various information like
accessibility modelling, evacuation modelling and the visualization of “best evacuation routes” from
each point of the study area.
Cooperation with the local administration and decision makers
From a very early stage of this research, local administration and decision makers were contacted
and every process step was agreed accordingly. Different workshops and meetings in Bali were
essential to intensify the cooperation with local stakeholders and to legitimate the field work on
site. The exchange of experiences and information regarding evacuation planning was an
important basis for the composition of this thesis. Based on the gained expertises during this work,
important stakeholders for evacuation planning could be identified.
The reference of achievements obtained in answering the research questions is presented in
Table 28.
96
No Research questions Reference
1. Research Objective: Definition of requirements for evacuation planning
What spatial data are needed for evacuation planning? Chapter 3
What are the most important needs for the end users (local stakeholders, exposed population)?
Chapter 3
2. Research Objective: Development and application of a spatial information system for tsunami evacuation planning
Based on literatures and expert interviews, which spatial information tools are suitable for evacuation modelling?
Chapter 3.4 and 5.5
What has to be considered in developing a spatial information tool to ensure its transferability to other coastal areas?
Chapter 6 and Chapter 5.6
Which spatial parameters have to be considered for tsunami evacuation modelling?
Chapter 3.1, 3.2 and 3.3
3. Research Objective: Giving a statement about the endangerment of the study area
How is the tsunami hazard locally distributed? Chapter 3.2 and 4.2
How many people are exposed? Chapter 3.3 and 4.3
In the context of evacuation planning, where are critical facilities?
Chapter 3.1.3 and 4.1
Where are evacuation bottlenecks? Chapter 5.5
4. Research Objective: Giving a decision support for evacuation in case of a tsunami event
Where are potential evacuation shelters? Chapter 3.1.3, 3.2.2, 4.1 and 4.2
Where are potential evacuation routes? Chapter 3.1.3, 3.4, 4.1, 4.4
Are the people in risk areas able to evacuate in a given period of time?
Chapter 3.5 and 4.5
Which spatial information products are useful for the local decision makers?
Chapter 4 and Chapter 5.5
5. Research Objective: Cooperation with the local administration and decision makers
Which stakeholders should be incorporated in the research process?
Chapter 5.6
How to transfer the gained knowledge and create acceptance for the local population?
Chapter 6
Table 28 Reference of achievements obtained in answering the research questions (Source: own composition)
97
Recommendations
The thesis has illustrated several fields were further research is recommended.
The great importance of vertical evacuation in urban areas is frequently mentioned. The results of
the evacuation modelling process presented in Chapter 4.5.2 show that the current defined ESB’ s
are not sufficient to cover the whole area. The “Service location planning” approach as discussed
in Chapter 5.5 can be used to optimize the current building distribution. While the designation of
an existing building as an ESB is strongly related with certain criteria, a technical inspection is
required. Therefore, the developed criteria check list for ESB (see Figure 42) provides extensive
information and recommendations both regarding general building requirements and current
building functions that can be assigned as ESB.
ESB designation or allocation is also an iterative process along with the decision making process of
spatial planning. An ESB can be built in the form of an evacuation-designated building (exclusively
shelter function), or can be a multi-function building such as a mosque, school and office as listed
before in Table 8. Considering general ESB requirements in the planning phase of such buildings
would be an important step to improve the evacuation ability in urban areas. In risk assessment
and management, the presence of ESB function will become a coping mechanism that reduces
people’s vulnerability. The dynamic and complex relationship among these components of risk,
such as how far can an ESB construction reduce the vulnerability, is also an interesting topic for
research.
A main objective of this thesis was to give a decision support in case of a tsunami event. Official
evacuation planning is a task for local authorities and therefore the gained results and expertises
have to be implemented in a local planning process. The identification of high risk areas, based on
the results of the accessibility and population modelling is essential for effective evacuation
planning while evacuation procedures can be focused accordingly. Critical bottlenecks can be
caused by a high population density in combination with a lack of evacuation possibilities or by the
concentration of high frequented evacuation routes. By the simulation of traffic flows using a tool
to identify the best evacuation route in the CWD concept, problems during an evacuation can be
identified early and – if possible – be mitigated accordingly. This information is also suitable to
improve the signing of potential evacuation routes. If the evacuation of a hazard zone is to
proceed in an orderly manner, it is essential that people know where to go, and what route to
take. The identified shelter catchment areas in the accessibility modelling are bordering areas
which are allocated to a certain evacuation shelter. To ensure a coordinated evacuation process,
evacuation signs have to guide the correct way. A combination of the tool to identify the best
evacuation route and the catchment area distribution provides a great possibility to bypass
evacuees from high frequented evacuation routes to alternative routes which also lead to the same
evacuation shelter.
98
A subsequent process of this work should intend the strengthening of the local capacities
regarding technical evacuation expertises. Under the direction of GTZ- IS, a first cooperation
process took place during the field work. Stakeholders which should be incorporated in the
research process were identified and common process steps were concerted. Acceptance and
usage of all operations in the local evacuation planning process have to be gained throughout
intensive participation of all stakeholders involved. To provide and support trainings for key
institutions to minimize knowledge gaps and to increase standard capacities should also be part of
the cooperation process in the sense of capacity building.
99
7 Outlook
The research for the study area was conducted to show prospects of supporting official evacuation
planning by the use of a GIS- based spatial information system. Decisive expertises are exposed in
the previous Chapters whereas some limitations of the approach are already addressed. The
tsunami hazard assessment providing estimates of areas that will be flooded is based on a single
inundation scenario and could be improved by using a multi- scenario approach in order to look at
the statistical probability that a certain tsunami event affects the study area. The calculation of the
exposed population is based on statistical data and estimations of other population groups as a
result of insufficient data. Due to the decisive dependence of the evacuation ability on detailed
population data, further data collection is required. A detailed quantification of the tourist
population and the daily commuters is necessary to gain a more reliable data base showing the
actual population in the study area. An assessment of social population vulnerability identifying
several characteristics that contribute to differentiate the ability to cope with a tsunami hazard,
can be used to highlight population groups that need special evacuation assistance. Also the
population disaggregation concept has to be improved while the weightings factors for the certain
functional zones are only based on expert interviews and field observation. Further literature
research regarding intra- destination movements of certain population groups would provide a
better conceptual basis for this approach.
Regarding the total approach, evacuation modelling based on the CWD concept meet the demands
of a spatial information system by identifying relevant spatial issues for evacuation planning. By
the integration of additional parameter, different analyses steps can be improved as presented in
the previous Chapters. The dynamic process of an evacuation is only considered by static
parameters. To visualize the temporal behaviour of population during an evacuation, modelling
software dedicated to analyzing and displaying interaction or flow data has to be integrated into
the concept. The Flowmap software can be mentioned exemplarily by providing useful measures
for accessibility analyses.
101
Bibliography
ADPC (2007A): EVACUATION ROUTES TOOLS ArcGIS toolbox - User’s manual.- Italian Ministry for
the Environment Land and Sea. Department for Environmental Research and Development,
Bangkok, 88 pp.
ADPC (2007B): CRATER – Short project description.- Italian Ministry for the Environment Land and
Sea. Department for Environmental Research and Development, Bangkok, 35 pp.
BAKOSURTANAL (ED.) (2003): Bali from space.- Indonesia Tourism Vol. 1, Cibinong Bogor, 160 pp.
BANGER, S.M. (2002): Remote sensing and geographical information system for natural disaster
management.- Geospatial application papers (Natural hazard management). Source:
http://www.gisdevelopment.net/application/natural_hazards/overview/nho0011.htm (viewed June
2008).
BATTY, M. (1994): Using GIS for visual simulation modelling.- In: GIS World 7(10), pp. 46 – 48.
BIRKMANN, J., FERNANDO, N., HETTIGE, S. ET AL. (EDS.) (2007): Rapid Vulnerability Assessment in Sri
Lanka.- SOURCE, Publication Series of UNU-EHS 7, Bonn, 84pp.
BUDIARJO, A. (2006): Evacuation Shelter Building Planning for Tsunami- prone Area; a Case Study
of Meulaboh City, Indonesia.- Master thesis, International Institute for Geo- information Science
and Earth Observation, Enschede, 112 pp.
BUNDESMINISTERIUM DES INNEREN (ED.)(2005): National plan for Information Infrastructure
Protection, Berlin, 21 pp.
BURKE, L. M. (1993): Race and environmental equity: a geographical analysis of Los Angeles.- In:
Geo Info Systems 3, pp. 44 - 50.
CASTELLANOS- ABELLA, E. (2002): GIS for Natural Disaster Management.- III Congreso Internacional
GEOMATICA 2002, 18-23 Febrero, La Habana, Cuba, 15 pp. Source: http://eacastellanos-
.googlepages.com/2002castellanos.pdf (viewed June 2008)
CHURCH, R.L. AND SEXTON. R.M. (2002): Modelling small area evacuation: Can existing transportation
infrastructure impede public safety?- Final Report, Vehicle Intelligence & Transportation Analysis
Laboratory, University of California, Santa Barbara, 22 pp.
COLE, J.W., SABEL, C.E., BLUMENTHAL, E., FINNIS, K., DANTAS, A., BARNARD, S. AND JOHNSTON, D.M.
(2005): GIS- based emergency and evacuation planning for volcanic hazards in New Zealand.- In:
Bulletin of the New Zealand society for earthquake engineering 38(3), pp.149 – 164.
COVA, T.J. AND CHURCH, R.L. (1997): Modelling community evacuation vulnerability using GIS. – In:
International Journal of Geographical Information Science 11(8), pp. 763 – 784.
DANGERMOND, J. (1991): Applications of GIS to the international decade for natural hazards
reduction.- In: Proceedings of the Fourth International Conference on Seismic Zonation, 26-29
August, Stanford University, California, pp. 445 - 468.
102
DE SILVA, F., PIDD, M. AND EGLESE, R. (1993): Spatial decision support systems for emergency
planning: an operational research/geographical information systems approach to evacuation
planning.- In: Proceedings of the 1993 Simulation Multiconference on the International Emergency
Management and Engineering Conference, 29 March-1 April, Arlington, San Diego, pp. 130 - 133.
DE SILVA, F. AND EGLESE, R. (2000): Integrating simulation modelling and GIS: spatial decision
support systems for evacuation planning.- In: Journal of the Operational Research Society 51, pp.
423 – 430.
EISNER, R. AND NTHMP (2001): Designing for Tsunamis. Seven principles for planning and designing
for tsunami hazards.- US National Tsunami Hazard Mitigation Program Steering Committee,
Sacramento, 60 pp.
EMERGENY MANAGEMENT AUSTRALIA (ED.) (2005): Evacuation Planning.- In: Australian Emergency
Manual Series 11, 30 pp.
FARROW, A. AND NELSON, A. (2001): Accessibility modelling in ArcView 3. An extension for computing
travel time and market catchment information.- Source: http://www.ciat.cgiar.org/access/pdf/ /ciat_access.pdf (viewed June 2008)
GATRELL, A.C. AND VINCENT, P. (1991): Managing natural and technological hazards.- In: Masser, I.
and Blakemore, M. (Eds.): Handling Geographical Information: Methodology and Potential
Applications, Longman, London, pp. 148 - 180.
GONZÁLEZ, F.I.,TITOV, V.V.,MOFJELD, H.O.,VENTURATO, A.J., SIMMONS, R.S., HANSEN, R., COMBELLICK, R.,
EISNER, R.K., HOIRUP, D.F., YANAGI, B.S.,YONG, S., DARIENZO, M., PRIEST, G.R., CRAWFORD, G.L. AND
WALSH, T.J. (2005): Progress in NTHMP Hazard Assessment.- In: Natural Hazards 35(1), pp 89-
110.
GTZ- IS (2007): Capacity Building in Local Communities. German- Indonesian Cooperation for
Tsunami Early Warning System.- Newsletter No.4, October- December, Eschborn, 12 pp.
GTZ- IS (2008): Capacity Building in Local Communities. German- Indonesian Cooperation for
Tsunami Early Warning System.- Newsletter No.5, January- March, Eschborn, 12 pp.
HOFSTEE, P. AND ISLAM, M. (2004): Disaggregation of census districts: better population information
for urban risk management.- In: ACRS (2004): Proceedings of the 25th Asian conference on remote
sensing, ACRS Silver jubilee, November 22-26, 2004, Chiang Mai, Thailand, Vol. 1 and 2, pp. 1206
– 1211.
JOHNSON, G.O. (1992): GIS applications in emergency management.- In: URISA Journal 4, pp. 66-
72.
JONG, T.D. AND ECK, J.R.V. (1996): Location profile- based measures as an improvement on
accessibility modelling in GIS.- In: Computers, Environment and Urban Systems 20(3), pp. 181 –
190.
103
KLÜPFEL, H., SCHRECKENBERG, M., MEYER-KÖNIG, T. (2004): Modelling for crowd movement and egress
simulation.- In: Hoogendoorn, S. et al. (Eds.): Traffic and Granular Flow ‘03, Springer, Berlin, pp.
357 – 372.
KHOMARUDIN, R., STRUNZ, G., POST, J., ZOßEDER, K., LUDWIG, R. (2008): Spatial improvement of
information on people density distribution using GIS approaches: an input for tsunami people
exposure assessment.- Abstract paper, National Conference on Tsunami Warning 2008, Bali,
Indonesia, 2 pp., unpublished.
LATIEF, H., PUSPITO, N.T. AND IMAMURA, F. (2000): Tsunami Catalogue and Zones in Indonesia.- In:
Journal of Natural Disaster Science 22(1), pp. 25 – 43.
LAU, G. AND MCKERCHER, B. (2006): Understanding tourist movement patterns in a destination: A
GIS approach.- In: Tourism and Hospitality Research 7(1), pp. 39 – 49.
MOSELEY, M.A. (1979): Accessibility: The rural challenge, London: Methuen.
NADLER, A. (2005): Bali – Reise Handbuch. Dormagen, 734 pp.
Pittmann, J., Pham, H., Court, M., Trafalis, T. and Ling, C. (2006): Dynamic Evacuation Modeling.-
Extended abstract, Conference on Dynamics of Disasters, October 5-7, Athens, Greece, 2pp.
POST, J., ZOßEDER, K., STRUNZ, G., BIRKMANN, J., GEBERT, N., SETIADI, N., ANWAR, H.Z., HARJONO, H.,
NUR, M. AND SIAGAN, T. (2007): Risk and Vulnerability assessment to tsunami and coastal hazards in
Indonesia: Conceptual framework and indicator development.- Proc. International Symposium on
disaster in Indonesia: Problem & Solution, 26-28 July, Padang, Indonesia, 10 pp.
REJESKI, D. (1993): GIS and risk: a three culture problem.- In: Goodchild, M.F., Parks, B.O.,
Steyaert, L.T. (Ed.): Environmental Modelling with GIS, Oxford University Press, New York, pp. 318
- 331.
SCHNEIDERBAUER, S. (2007): Risk and Vulnerability to Natural Disasters - from Broad View to
Focused Perspective.- Dissertation, Faculty for Geoscience, University of Berlin, 120 pp.
SLEETER, R. AND WOOD, N. (2006): Estimating daytime and nighttime population density for coastal
communities in Oregon.- Urban and regional information systems association, Proceedings of the
Annual Conference, September 26-29, Vancouver, Canada, 15 pp.
SUGIMOTO, T., MURAKAMI, H., KOZUKI, Y. AND NISHIKAWA, K. (2003): A human damage prediction
method for tsunami disaster incorporating evacuation activities.- In: Natural Hazards 29, pp 585 -
600.
TAUBENBÖCK, H., ROTH, A. AND DECH, S. (2007): Linking structural urban characteristics derived from
high resolution satellite data to population distribution.- In: Coors, V., Rumor, M., Fendel, E. and
Zlatanova , S. (Eds.): Urban and Regional Data Management, Taylor & Francis Group, London, pp.
35 – 45.
104
TAUBENBÖCK, H., POST, J., ROTH, A., ZOßEDER, K., STRUNZ, G., DECH, S. (2008): A conceptual
vulnerability and risk framework as outline to identify capabilities of remote sensing.- In: Natural
Hazards and Earth System Sciences 8, pp. 409 – 420.
STRUNZ, G., POST, J., ZOßEDER, K. (2008A): Draft guidelines hazard assessment. UNESCO IOC ICG
WG 3: Risk assessment, Version 0.5.- Working paper, German Remote Sensing Data Center (DFD),
unpublished.
STRUNZ, G., POST, J., ZOßEDER, K. (2008B): Draft guidelines exposure component. UNESCO IOC ICG
WG 3: Risk assessment, Version 0.5.- Working paper, German Remote Sensing Data Center (DFD),
unpublished.
VORLAUFER, K. (1999): Bali - Massentourismus und nachhaltige Entwicklung: Die sozio-
ökonomische Dimension.- In: Erdkunde 53(4), pp. 273 – 301.
UNESCO- IOC (2006): Tsunami Glossary.- IOC Information Document No.1221, 36 pp.
ZWAN, J.V.D., WEL, R.V.D., JONG, T.D. AND FLOOR, H. (2005): Manual Flowmap 7.2.- Faculty of
Geographical Sciences, University of Utrecht, Utrecht, 160 pp.
Expert interviews
DR. JOACHIM POST, German Remote Sensing Data Center (DFD) of DLR, Department on
Environment and Security, Oberpfaffenhofen – 15.04.2008.
DR. KAI ZOßEDER, German Remote Sensing Data Center (DFD) of DLR, Department on Environment
and Security, Oberpfaffenhofen – 15.04.2008.
DR. KAI ZOßEDER, German Remote Sensing Data Center (DFD) of DLR, Department on Environment
and Security, Oberpfaffenhofen – 23.04.2008.
DR. JOACHIM POST, German Remote Sensing Data Center (DFD) of DLR, Department on
Environment and Security, Oberpfaffenhofen – 19.05.2008.
105
Appendices
Appendix 1 Standardized questionnaire for the hotel sector in the study area (Source: own composition)
106
Appendix 2 Standardized questionnaire for the health sector in the study area (Source: own composition)
107
Appendix 3 Standardized questionnaire for the school sector in the study area (Source: own composition)
108
Appendix 4 Evacuation shelter accessibility within 25 min, calculated for day and nighttime with Method B (own illustration – Data basis: CWD surface; Map basis: catchment areas of the evacuation shelters, population data from Method B)
109
Appendix 5 Evacuation shelter accessibility within 25 min, calculated for day and nighttime with Method A (own illustration – Data basis: CWD surface; Map basis: catchment areas of the evacuation shelters, population data from Method A)